CN105527968A - Information processing method and information processing device - Google Patents

Abstract

The invention discloses an information processing method and an information processing device. The method is applied to mobile electronic equipment for map construction. The method comprises steps that, a present common node is generated according to a frame of presently-acquired observation data and present pose information of the mobile electronic equipment; whether the frame of observation data contained in the present common node is in matching with the integrated observation data contained in a previous crucial node is determined, and the integrated observation data is acquired through integrating multiple frames of observation data contained in multiple common nodes; when the frame of observation data contained in the present common node is in matching with the integrated observation data contained in the previous crucial node, pose information of all nodes is adjusted according to the present pose information contained in the present common node; all crucial nodes are reconstructed according to the pose information of the all nodes after adjustment; the global map is updated according to all the crucial nodes after reconstructed.

Description

Information processing method and signal conditioning package

Technical field

The present invention relates to field of computer technology, more specifically, the present invention relates to a kind of information processing method and signal conditioning package.

Background technology

Instant location and map structuring (SimultaneousLocalizationandMapping, SLAM) are the popular research topics at present in robot localization.So-called SLAM is exactly by mobile electronic equipment (such as, mobile robot) location creates combine together with environmental map, namely robot builds increment type environmental map according to the perception of the estimation of self pose and sensors towards ambient in motion process, utilizes this map to realize the location of self simultaneously.

SLAM method based on topological diagram is a kind of conventional SLAM method, and it can be detected by closed circuit and scheme to optimize the cumulative movement error reducing robot and produce.Particularly, this SLAM method uses topological diagram to represent the operating path of mobile robot.In topological diagram, a pose of each node on behalf robot, and the limit between these nodes comprises conversion and confidence information.Such as, a new node can be created, to present a new robot pose, and add a limit between this node and its previous adjacent node, to introduce sensor measurement, thus two nodes that constraint is connected.Finally, after closed circuit being detected between the two nodes and with the addition of the limit of closed circuit, this SLAM method can optimize by figure the cumulative movement error reducing robot.

But, in this SLAM method, because the observation data comprised in each node is very limited, so when closed circuit detects, probably there is the error hiding between observation data, thus cause the accuracy of closed loop detect may be very low.

In order to detect closed loop better, a kind of SLAM method based on topological diagram of improvement proposes, and can increase the SLAM flow process of a rank (level) in existing SLAM flow process again.Particularly, (namely the SLAM method of this improvement except being used in the common node that defines in original SLAM method, more low-level node) represent robot pose outside, also when multiple ordinary node meets predetermined condition, by merge observation data by described multiple ordinary node stipulations be super node (namely, the node of higher level), delete these ordinary nodes simultaneously, and in higher level, build subgraph (submap) based on this super node, thus realize loop detection and the map fusion of global map.In this SLAM method, because the observation data in each super node is abundanter, so improve the accuracy of closed loop detect.

But, in practice, due to the continuous advance along with robot, kinematic error in each ordinary node is in continuous increase, so the meromixis observation data comprised at the super node generated based on the ordinary node comprising error may also exist inconsistent frame matching, therefore, present if maintain the super node that there is deviation in higher level to carry out map, obviously, this error will cannot be eliminated forever.

Summary of the invention

In order to solve the problems of the technologies described above, according to an aspect of the present invention, provide a kind of information processing method, described method is applied to mobile electronic equipment, described mobile electronic equipment is used to carry out map structuring, and described method comprises: generate current ordinary node according to a frame observation data of current acquisition and the current posture information of described mobile electronic equipment; Judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at previous key node, wherein, described integration observation data is by integrating obtained to the multiframe observation data comprised at multiple ordinary node; When judging that the frame observation data comprised at described current ordinary node is mated with the integration observation data comprised at described previous key node, adjust the posture information of all nodes according to the current posture information comprised at described current ordinary node; Posture information according to all nodes after adjustment reconstructs all key nodes; And upgrade global map according to all key nodes after reconstruct.

In addition, according to a further aspect in the invention, provide a kind of signal conditioning package, described application of installation is in mobile electronic equipment, described mobile electronic equipment is used to carry out map structuring, described device comprises: ordinary node generation unit, for generating current ordinary node according to a frame observation data of current acquisition and the current posture information of described mobile electronic equipment; Data Matching judging unit, for judging whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at previous key node, wherein, described integration observation data is by integrating obtained to the multiframe observation data comprised at multiple ordinary node; Posture information adjustment unit, for when judging that the frame observation data comprised at described current ordinary node is mated with the integration observation data comprised at described previous key node, adjust the posture information of all nodes according to the current posture information comprised at described current ordinary node; Key node reconfiguration unit, for reconstructing all key nodes according to the posture information of all nodes after adjustment; And global map updating block, for upgrading global map according to all key nodes after reconstruct.

Compared with prior art, adopt the information processing method according to the embodiment of the present invention and device, can while generating key node according at least one ordinary node, retain this at least one ordinary node, and when judging that the frame observation data comprised at current ordinary node is mated with the integration observation data comprised at previous key node, the posture information of all nodes is adjusted according to the current posture information comprised at described current ordinary node, all key nodes are reconstructed according to the posture information of all ordinary nodes after adjustment and the posture information of all key nodes, thus upgrade global map according to all key nodes after reconstruct.Therefore, in an embodiment of the present invention, because key node comprises abundant picture information, thus can more robust and accurately detection loop close; And ordinary node can be regarded as when figure optimizes due to key node and participate in optimizing to reconstruct picture information, so when carrying out map in current in higher level, eliminate the inconsistent frame matching existed in meromixis observation data well.

Other features and advantages of the present invention will be set forth in the following description, and, partly become apparent from instructions, or understand by implementing the present invention.Object of the present invention and other advantages realize by structure specifically noted in instructions, claims and accompanying drawing and obtain.

Accompanying drawing explanation

Accompanying drawing is used to provide a further understanding of the present invention, and forms a part for instructions, together with embodiments of the present invention for explaining the present invention, is not construed as limiting the invention.In the accompanying drawings:

Fig. 1 illustrates the information processing method according to the embodiment of the present invention.

Fig. 2 illustrates the information processing method according to the concrete example of the embodiment of the present invention.

Fig. 3 illustrates the movement locus of the mobile electronic equipment according to the concrete example of the embodiment of the present invention.

Fig. 4 illustrates the more low-level topological diagram according to the concrete example of the embodiment of the present invention.

Fig. 5 A to Fig. 5 F illustrates the comprehensive topological diagram according to the concrete example of the embodiment of the present invention.

Fig. 6 illustrates the local search algorithm for closed loop detect according to the concrete example of the embodiment of the present invention.

Fig. 7 illustrates and calculates ratio juris according to the loop detection based on interpolation method of the concrete example of the embodiment of the present invention.

Fig. 8 illustrates and adjusts failed scene according to the pose that may cause of the concrete example of the embodiment of the present invention.

Fig. 9 illustrates the principle of the pose adjustment algorithm based on interpolation method according to the concrete example of the embodiment of the present invention.

Figure 10 A illustrates the design sketch of the map generated according to the SLAM method of prior art.

Figure 10 B illustrates the design sketch of the map generated according to the SLAM method of the embodiment of the present invention.

Figure 10 C illustrates the design sketch of the map generated according to the SLAM method of prior art.

Figure 10 D illustrates the design sketch of the map generated according to the SLAM method of the embodiment of the present invention.

Figure 10 E illustrates the design sketch of the map generated according to the SLAM method of prior art.

Figure 10 F illustrates the design sketch of the map generated according to the SLAM method of the embodiment of the present invention.

Figure 11 illustrates according to signal conditioning package of the present invention.

Figure 12 illustrates the mobile electronic equipment according to the embodiment of the present invention.

Embodiment

Describe in detail with reference to the accompanying drawings according to each embodiment of the present invention.Here it is to be noted that it in the accompanying drawings, identical Reference numeral is given there is identical or similar structures and function ingredient substantially, and the repeated description of will omit about them.

Fig. 1 illustrates the information processing method according to the embodiment of the present invention.

Information processing method shown in Fig. 1 can be applied to mobile electronic equipment, and described mobile electronic equipment is used to carry out map structuring.

In one embodiment, described mobile electronic equipment can be robot (Robot), for moving in targeted environment, simultaneously by performing instant location with map structuring (SLAM) to self carrying out quick position and generating the diagram data exactly about this targeted environment.

In one example, described mobile electronic equipment can independently move in targeted environment, performs instant location and map structuring (SLAM) simultaneously.

In another example, described mobile electronic equipment also can communicate with region server, move in targeted environment with the map datum provided according to region server and Route Planning Data, perform instant location and map structuring (SLAM) simultaneously, and send the result of described instant location and map structuring to region server, for upgrading the map datum stored in described region server.

As illustrated in figure 1, described information processing method comprises:

In step s 110, current ordinary node is generated according to a frame observation data of current acquisition and the current posture information of described mobile electronic equipment.

In one embodiment, after the current ordinary node of generation, described information processing method can also comprise: when the multiple ordinary nodes judging to comprise described current ordinary node meet key node formation condition, from the multiple ordinary nodes comprising described current ordinary node, select an ordinary node; Posture information according to described multiple ordinary node is integrated the multiframe observation data comprised at described multiple ordinary node, to obtain integrating observation data; And use described integration observation data to replace the frame observation data comprised at selected ordinary node, thus selected ordinary node is converted to current key node.

In the step s 120, judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at previous key node, wherein, described integration observation data is by integrating obtained to the multiframe observation data comprised at multiple ordinary node.

In one embodiment, before whether the frame observation data judging to comprise at described current ordinary node mates with the integration observation data comprised at previous key node, described information processing method can also comprise: judge whether meet matching detection condition between described current ordinary node and described previous key node; And when judging to meet described matching detection condition between described current ordinary node and described previous key node, read in the integration observation data that described previous key node comprises.

In one example, judge that whether meeting matching detection condition between described current ordinary node and described previous key node can comprise: calculate the spacing distance between described current ordinary node and described previous key node according to the posture information of described current ordinary node and the posture information of described previous key node; More described spacing distance and predetermined threshold; When described spacing distance is less than or equal to described predetermined threshold, judge to meet described matching detection condition between described current ordinary node and described previous key node; And when described spacing distance is greater than described predetermined threshold, judge not meet described matching detection condition between described current ordinary node and described previous key node.

In one embodiment, judge whether the frame observation data comprised at described current ordinary node is mated can comprise with the integration observation data comprised at previous key node: generate multiple initial estimate according to the posture information of described current ordinary node and the posture information of described previous key node; And utilize described multiple initial estimate, and use error equation, judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at described previous key node.

In one example, generate multiple initial estimate according to the posture information of described current ordinary node and the posture information of described previous key node can comprise: determine the spacing distance between described current ordinary node and described previous key node according to the posture information of described current ordinary node and the posture information of described previous key node; And between predetermined value and described spacing distance, select multiple value, as described multiple initial estimate.

In one example, between predetermined value and described spacing distance, multiple value is selected to comprise: between described predetermined value and described spacing distance, select N number of value according to equal intervals,

Wherein, N is the number of initial estimate, be capping function, Dis is described spacing distance, and T _loopthe convergence in mean being described error equation is interval.

In one example, utilize described multiple initial estimate, and use error equation, judge whether the frame observation data comprised at described current ordinary node is mated can comprise with the integration observation data comprised at described previous key node: the error range determining described current ordinary node, described error range indicates the error between posture information and the attained pose of described mobile electronic equipment comprised at described current ordinary node; The initial estimate be in described error range is selected, alternatively estimated value among described multiple initial estimate; And only utilize described candidate's estimated value, and use error equation, judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at described previous key node.

In step s 130, which, when judging that the frame observation data comprised at described current ordinary node is mated with the integration observation data comprised at described previous key node, adjust the posture information of all nodes according to the current posture information comprised at described current ordinary node.

In one embodiment, the posture information adjusting all nodes according to the current posture information comprised at described current ordinary node can comprise: select each node as adjustment aim node; The posture information of described adjustment aim node is redefined according to the current posture information comprised at described current ordinary node; And the original posture information of described adjustment aim node is at least adjusted according to the posture information redefined.

In one example, the original posture information that the posture information that at least basis redefines adjusts described adjustment aim node can comprise: the original posture information according to described adjustment aim node generates multiple initial estimate with the posture information redefined; Utilize described multiple initial estimate, and use error equation, determine final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum; And use described final estimated value as object pose information to adjust the original posture information of described adjustment aim node.

In one example, utilize described multiple initial estimate, and use error equation, determine final estimated value, to make, the matching error between described adjustment aim node and adjacent node is minimum can be comprised: the error range determining described adjustment aim node, and described error range indicates the error between posture information and the attained pose of described mobile electronic equipment comprised at described adjustment aim node; The initial estimate be in described error range is selected, alternatively estimated value among described multiple initial estimate; And only utilize described candidate's estimated value, and use error equation, determine final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum.

In step S140, the posture information according to all nodes after adjustment reconstructs all key nodes.

In one embodiment, reconstruct all key nodes according to the posture information of all nodes after adjustment can comprise: for each key node, posture information according to all ordinary nodes after the adjustment comprised at described key node is reintegrated the multiframe observation data comprised at described all ordinary nodes, to obtain the integration observation data after upgrading; And use the integration observation data after described renewal to replace original integration observation data, thus reconstruct described key node.

In step S150, upgrade global map according to all key nodes after reconstruct.

As can be seen here, adopt the information processing method according to the embodiment of the present invention, can while generating key node according at least one ordinary node, retain this at least one ordinary node, and when judging that the frame observation data comprised at current ordinary node is mated with the integration observation data comprised at previous key node, the posture information of all nodes is adjusted according to the current posture information comprised at described current ordinary node, all key nodes are reconstructed according to the posture information of all ordinary nodes after adjustment and the posture information of all key nodes, thus upgrade global map according to all key nodes after reconstruct.Therefore, in an embodiment of the present invention, because key node comprises abundant picture information, thus can more robust and accurately detection loop close; And ordinary node can be regarded as when figure optimizes due to key node and participate in optimizing to reconstruct picture information, so when carrying out map in current in higher level, eliminate the inconsistent frame matching existed in meromixis observation data well.

Hereinafter, the concrete example of the information processing method according to the embodiment of the present application is described with reference to Fig. 2 to Fig. 9.

Fig. 2 illustrates the information processing method according to the concrete example of the embodiment of the present invention.

In the concrete example of the embodiment of the present application, to be described in following application scenarios, in this application scenarios, information processing method according to the concrete example of the embodiment of the present application can be implemented in robot, this robot may be used for moving in circumstances not known, simultaneously by performing instant location with map structuring (SLAM) to self carrying out quick position and generating the diagram data exactly about this circumstances not known.

It should be noted that, the present invention is not limited thereto.Such as, the information processing method according to the concrete example of the embodiment of the present application also can be implemented in other electronic equipments, as long as this electronic equipment or can depend on sports equipment and carry out moving in circumstances not known by means of the kinematic system of self.In addition, also may be used for carrying out map structuring to known environment according to the information processing method of the concrete example of the embodiment of the present application.

As illustrated in Figure 2, described information processing method comprises:

In step S210, generate current ordinary node.

When needs utilize mobile electronic equipment (such as, movable machine people Robot) come to one circumstances not known perform explore time, can, by region server in advance for Robot fictionalizes a collisionless optimal path, robot can be moved in circumstances not known according to this optimal path.Alternatively, this robot also can independently move in circumstances not known, to detect place environment and to understand.Then, this robot, while carrying out moving, performs instant location and map structuring (SLAM) to this circumstances not known.

Along with mobile robot carries out moving and taking in circumstances not known, robot can be triggered based on various different trigger condition and constantly obtain observation data and the posture information of self at each location point.

In one example, robot can be triggered based on predetermined time interval (such as, every 10 seconds) and perform data acquisition operations at different location points.

But, obviously the present invention is not limited thereto.Such as, also can other perform data acquisition operations at different location points because of usually trigger unit device people based on the space distribution of predetermined travel distance (such as, every 10 meters), unique point etc.Or, scan request can also be sent to robot, make this robot only when receiving scan request, just performing data acquisition operations.

Then, between the moving period of robot, it is constantly performing SLAM operation based on fixed time interval at each location point, and obtains result data.

Such as, first this robot can generate a current ordinary node according to the frame observation data got at current location point and the current posture information of self.

Fig. 3 illustrates the movement locus of the mobile electronic equipment according to the concrete example of the embodiment of the present invention, and Fig. 4 illustrates the more low-level topological diagram according to the concrete example of the embodiment of the present invention.

In the concrete example of the embodiment of the present invention, suppose that robot takes regular-shape motion track to move in circumstances not known.As illustrated in fig. 3, robot from start position 1 setting in motion to position 2, then continues to move to position 3, so continues, and then moves to position 11, and finally get back to start position 1, thus complete the movement locus of rectangle.

Although it should be noted that and be here described the embodiment of the present invention for rectangular path, the present invention is not limited thereto.But robot can adopt arbitrary rule or irregular track and move in circumstances not known.

Particularly, for the position 1 of starting point, first robot performs map structuring at position 1 place.Such as, this robot obtains a frame observation data by data acquisition facility, and generates an ordinary node according to the posture information of self, i.e. ordinary node 1 as illustrated in figure 4, as current ordinary node.

Here, why it being referred to as ordinary node, is because it is a relative concept relative to key node.Such as, in this information processing method, key node generates by carrying out stipulations at least one ordinary node, and thus ordinary node is the more low-level node of one, and key node is a kind of node of higher level.

It should be noted that, in the diagram, for the ease of understanding, illustrate only more low-level ordinary node, and the key node of the higher level generated through stipulations is not shown.

Such as, the frame observation data reflection results that can be obtained by the laser transmitter projects laser equipped in robot.When 2D-SLAM, this frame observation data will comprise the two-dimensional coordinate (x, y) of each analyzing spot causing laser to return.Be 180 degree at the emission angle of laser and every launching beam of laser for 1 degree, the two-dimensional coordinate of 180 analyzing spots can be got, and form this frame observation data.

Obviously, although be described using laser scanning result as observation data above, the present invention is not limited thereto.Such as, when be equipped with camera in robot, this frame observation data also can be the view data that robot is captured by camera.Further, if this camera is the camera that can catch disparity map, then this frame observation data also can be the disparity map in surveyed area, so just can realize 3D-SLAM.Such as, this disparity map can be directly taken by special parallax video camera.Alternatively, also can pass through binocular camera, many orders camera, stereoscopic camera shooting gray-scale map (or cromogram), and then calculate corresponding disparity map according to described gray-scale map (or cromogram).In addition, disparity map is here not limited to just to be obtained by multiple camera, but also can be obtained based on time domain by a camera.Such as, piece image can be obtained as left image a moment shooting by a camera, then at subsequent time, shooting after this camera slightly shift position is obtained another piece image as right image, also can calculate disparity map based on the left image so obtained and right image.

In addition, such as, the posture information of robot can comprise the positional information (such as, coordinate information) (x, y) of this robot at position 1 place, and can comprise this robot at position 1 place towards angle information (θ).If suppose that this position 1 is the origin position of whole circumstances not known, then can learn that the coordinate figure of this present node is (0,0).

Then, the ordinary node of current generation can be stored in memory, for using after a while.

In step S220, current ordinary node and previous ordinary node are carried out frame matching.

After generating current ordinary node, the first matching process can be used, frame matching (scanmatch) be carried out to the last ordinary node of current ordinary node and described current ordinary node, thus continues to realize map structuring.

Such as, this first matching process can be iterative closest point (ICP) algorithm, limit scan matching (PSM) algorithm or other similar algorithms.

As shown in Figure 3 and Figure 4, due to this current ordinary node (namely, ordinary node 1) be starting point in robot motion's track, that is, this robot not yet produces any motion, and before it, there is not any history node yet, so now, can without the need to performing the operation of above-mentioned frame matching.Like this, be the above-mentioned hypothesis of the origin position of whole circumstances not known based on primary importance, can directly the coordinate figure of this ordinary node 1 in map be defined as (0,0).Or, more generally, if this primary importance is not set as origin position, also directly the coordinate figure of this ordinary node 1 in map can be defined as (x1, y1) based on initial setting.

In step S230, judge whether the multiple ordinary nodes comprising current ordinary node meet key node formation condition.

Next, can judge whether this current ordinary node and the previous ordinary node (or being referred to as, history ordinary node) stored in memory meet key node formation condition.If the multiple ordinary nodes comprising current ordinary node meet key node formation condition, then this information processing method may be advanced to step S240, with by described multiple ordinary node stipulations for key node.Otherwise this information processing method turns back to step S210 to be continued to perform, to obtain a new frame observation data, and so repeatedly, until robot no longer moves, that is, until again cannot get a new frame observation data.

Can trigger at least one ordinary node stipulations based on various different key node formation condition is key node.

Such as, this key node formation condition can trigger generation key node based on the range ability of robot or course length (such as, whenever robot travels 25 meters).Use course length to generate trigger condition as key node and can ensure higher grid map size, and can as the clue for detection loop.

It should be noted that, although be here described the embodiment of the present invention to generate key node based on distance, the present invention is not limited thereto.In one example, also generation key node can be triggered based on the number being not the residue ordinary node (comprising history ordinary node and current ordinary node) of key node by stipulations of preserving in memory.Such as, after last time generates key node to distance, when storing again 4 ordinary nodes in memory, be just key node by these 4 ordinary node stipulations.In another example, can also based on residue ordinary node in unique point space structure, observation area etc. other because of usually trigger generation key node.

As shown in Figure 3 and Figure 4, due to now, this current ordinary node (that is, ordinary node 1) is the starting point in robot motion's track, so can judge this ordinary node not movement exceed preset distance, thus do not meet key node formation condition.

So as mentioned above, this information processing method turns back to step S210 to be continued to perform.

Now, robot continues to move along the rectangular path of self, and when interval (such as, 10 seconds) after a predetermined time, continues to obtain the observation data in the next position (that is, position 2) by laser sensor.

Particularly, robot can continue at position 2 place to perform map structuring.

Such as, in step S210, this robot obtains next frame observation data (such as by data acquisition facility, the two-dimensional coordinate of 180 analyzing spots) and the current posture information (x of this robot, y, θ), to generate current ordinary node, i.e. ordinary node 2 as illustrated in figure 4, and store this ordinary node 2 in memory.

Then, in step S220, current ordinary node (that is, ordinary node 2) and previous ordinary node (that is, ordinary node 1) can be carried out frame matching.In topological diagram, the frame matching between this node can be expressed as intuitively and carry out company limit between the two nodes.

Such as, can first use such as iterative closest point (ICP) algorithm to estimate the transition matrix between ordinary node 2 and ordinary node 1.Then, the analyzing spot of some in ordinary node 2 can be determined, under these analyzing spots under the coordinate system of ordinary node 2 being transformed into the coordinate system of ordinary node 1 by this transition matrix, find at their nearest analyzing spots of ordinary node 1 middle distance, estimate the error between them, and again determine the analyzing spot of some, repeat said process, thus make this error minimize by the mode of iteration, until the accurate scanning realized between ordinary node 2 with ordinary node 1 is aimed at, to obtain the coordinate information (x2 of this ordinary node 2 in map, y2).

Next, in step S230, can judge whether this current ordinary node and the history ordinary node stored in memory meet key node formation condition based on distance.Suppose that now this current ordinary node (namely, ordinary node 2) starting point in distance robot motion track is 10 meters, so can judge this ordinary node 2 not movement exceed preset distance (such as, 25 meters), thus do not meet key node formation condition.

So, as mentioned above, this information processing method returns execution, to generate in step S210 and to store ordinary node 3, in step S220 by current ordinary node (that is, ordinary node 3) and previous ordinary node (namely, ordinary node 2) carry out frame matching, and judge that in step S230 this ordinary node 3 only moved 22 meters, still movement does not exceed preset distance, thus this information processing method returns execution again.

At this moment, generate in step S210 and store ordinary node 4, in step S220, current ordinary node (that is, ordinary node 4) and previous ordinary node (that is, ordinary node 3) being carried out frame matching.Suppose that now this current ordinary node (that is, ordinary node 4) is 30 meters apart from the starting point in robot motion's track, so can judge that this ordinary node 4 movement exceeds preset distance, thus meets key node formation condition.

At this moment, this information processing method proceeds to step S240.

In step S240, generate current key node.

Judging to comprise current ordinary node (namely, ordinary node 4) at interior multiple ordinary nodes (namely, ordinary node 1-4) meet key node formation condition after, an ordinary node can be selected from these ordinary nodes (that is, ordinary node 1-4).This selection course can be carried out at random, also can be artificial setting.

Such as, can from preserve in memory not by stipulations be key node residue ordinary node (that is, ordinary node 1-4) among select the ordinary node (that is, ordinary node 1) that generates at first, as datum node.

Then, can integrate the multiframe observation data comprised at these ordinary nodes according to the posture information of described multiple ordinary node, to obtain integrating observation data.

Such as, first, the observation information of the repetition comprised at all ordinary nodes (that is, ordinary node 1-4) can be filtered.

Particularly, can characteristic information extraction (feature) and cartographic information the observation data (observation) from ordinary node 1 to ordinary node 4, judge whether there is duplicate message between these information.If so, then can filter the characteristic information repeated and cartographic information, to generate integration observation data, thus the expense decreased when performing SLAM compared with low level and redundant information.

Next, after the characteristic information filtering out the repetition that current ordinary node and history ordinary node comprise in compared with low level, integration observation data can be used to replace the original observation data comprised in ordinary node 1, thus selected ordinary node 1 is converted to current key node 1.

Preferably, in the process of carrying out characteristic filter, the operation of enhancing Map recognition can also be carried out to each characteristic information, to reduce the error that may exist in characteristic information, and improve the precision of Map recognition.

Particularly, the characteristic information of the repetition that can comprise to current ordinary node and history ordinary node gives weighted value respectively, and such as, the allocation rule of weighted value can be: if robot is nearer apart from the distance of this unique point, then weighted value is larger.Here, suppose that this unique point is the potted flower that robot photographs in traveling process, and robot is this potted flower of ordinary node 1 middle distance 9 meters, this potted flower of ordinary node 2 middle distance 6 meters, this potted flower of ordinary node 3 middle distance 3 meters, then such as, lower weighted value (such as, 2) can be given to ordinary node 1, medium weighted value is given (such as to ordinary node 2,, and give higher weights value (such as, 5) to ordinary node 3 3).Then, can be weighted on average according to the characteristic information of described weighted value to the repetition that multiple node comprises, and generate the integration observation data of the characteristic information after comprising weighted mean, to generate single key node.

Then, the key node 1 of current generation can be stored in memory, for using after a while.

It should be noted that, key node and other ordinary node basic simlarity, its difference is only, key node further comprises the grid map information (or being referred to as subgraph) in a regional area, and ordinary node only comprises the grid map information in the overall situation.The motivation building key node is, can be undertaken mating instead of carrying out coupling to judge whether occurring closed circuit by single frames observation data by Local Subgraphs, thus increases the accuracy of loop detection.

Like this, in the process of map structuring, in more low-level ordinary node, global grid map can be constructed; And in the key node of higher level, global grid map and local grid map can be constructed simultaneously.Safeguard this global grid map, and apply Markov Monte Carlo (MCMC) theory to estimate the position of robot in world coordinate system; And safeguard this Local grid map, whether there is closed circuit for checking.

In step s 250, judge whether current ordinary node and previous key node meet matching detection condition.

After generation current key node (that is, key node 1), can judge whether current ordinary node and previous key node meet the condition of matching detection (that is, closed circuit detects).If current ordinary node and previous key node meet matching detection condition, then this information processing method may be advanced to step S260, detects to carry out matching detection (that is, closed circuit).Otherwise this information processing method turns back to step S210 to be continued to perform, to obtain a new frame observation data.

The detection operation of closed circuit can be triggered based on various different testing conditions.Such as, closed circuit testing conditions can carry out detection trigger operation based on the distance between two nodes.Particularly, if find that current ordinary node and previous key node are within preset distance, then think that two nodes meet closed circuit testing conditions.That is, when the current pose of robot is close to previous key node, can think that two nodes meet closed circuit testing conditions.

As shown in Figure 3 and Figure 4, due to now, this current ordinary node (that is, ordinary node 4) belongs to current key node (that is, key node 1), so can judge that this ordinary node does not meet closed circuit testing conditions.

So, as mentioned above, this information processing method turns back to step S210 to be continued to perform, then in turn four ordinary nodes are generated and stored (namely in position 5 to the observation data of position 8 according to robot, ordinary node 5-8), frame matching is carried out to every two adjacent ordinary nodes; Judge whether to meet key node formation condition, thus by carrying out integration to the observation data of ordinary node 5-8, ordinary node 5 is converted to key node 2, and judge whether whether ordinary node 8 meet matching detection condition with key node 1.As shown in Figure 3 and Figure 4, obviously, when robot is in position 8, posture information according to current ordinary node 8 and previous key node 1 can judge that spacing distance between the two is obviously greater than predetermined threshold (such as, 5 meters), that is, there is not the possibility producing closed circuit between the two, therefore, still judge that this ordinary node 8 does not meet matching detection condition with key node 1.

So analogize, along with robot moves to position 11 through position 10 from position 9, then generate and store again three ordinary nodes (that is, ordinary node 9-11), frame matching is carried out to every two adjacent ordinary nodes; Judge whether to meet key node formation condition, thus by carrying out integration to the observation data of ordinary node 9-11, ordinary node 9 is converted to key node 3, and judge whether whether ordinary node 11 meet matching detection condition with key node 1 and key node 2.As shown in Figure 3 and Figure 4, obviously, when robot is when running to position 11, posture information according to current ordinary node 11 and previous key node 1 can judge that spacing distance between the two has been less than predetermined threshold (such as, 5 meters), that is, the possibility occurring there is closed circuit can be known between the two, therefore, can judge because this ordinary node 11 and key node 1 are apart from very near, so met matching detection condition.

Before continuing to describe follow-up step S260, in order to the generative process of each node above-mentioned in topological diagram is clearly described further, below, the process that comprises the topological diagram of ordinary node and key node is generated incrementally in being described in detail according to the embodiment of the present invention concrete example with reference to figure 5A to Fig. 5 F.

Fig. 5 A to Fig. 5 F illustrates the comprehensive topological diagram according to the concrete example of the embodiment of the present invention.In this comprehensive topological diagram, both comprise more low-level ordinary node, also comprise the key node of the higher level generated through stipulations.

In Fig. 5 A to Fig. 5 F, white hollow circle represents ordinary node, and solid black circle represents key node, and grey filling circle represents the current node processed.Each key node comprises the observation information in several adjacent ordinary nodes, thus forms the subgraph of local.

As shown in Figure 5A, by repeating step S210 to S230, in this topological diagram, four ordinary node 1-4 are generated.Next, owing to judging to comprise current ordinary node (namely in step S230, ordinary node 4) multiple ordinary nodes (namely, ordinary node 1-4) meet key node formation condition, so in step S240, by integrating the observation data in ordinary node 1-4, thus ordinary node 1 is converted to key node 1, this key node 1 indicates Local Subgraphs 1 (being represented by M1 in the drawings), as shown in Figure 5 B.Then, owing to judging there is no previous key node in step s 250, continue to perform so this information processing method returns.

As shown in Figure 5 C, by repeating step S210 to S230, in this topological diagram, four ordinary node 5-8 are generated again.Next, owing to judging to comprise current ordinary node (namely in step S230, ordinary node 8) multiple ordinary nodes (namely, ordinary node 5-8) meet key node formation condition, so in step S240, by integrating the observation data in ordinary node 5-8, thus ordinary node 5 is converted to key node 2, this key node 2 indicates Local Subgraphs 2 (being represented by M2 in the drawings), as shown in Figure 5 D.Then, owing to judging that current ordinary node (that is, ordinary node 8) and previous key node (that is, key node 1) do not meet matching detection condition in step s 250, continue to perform so this information processing method returns.

As shown in fig. 5e, by repeating step S210 to S230, in this topological diagram, three ordinary node 9-11 are generated again.Next, owing to judging to comprise current ordinary node (namely in step S230, ordinary node 8) multiple ordinary nodes (namely, ordinary node 9-11) meet key node formation condition, so in step S240, by integrating the observation data in ordinary node 9-11, thus ordinary node 9 is converted to key node 3, this key node 3 indicates Local Subgraphs 3 (being represented by M3 in the drawings), as illustrated in figure 5f.Then, can be in step s 250, judge current ordinary node (namely, ordinary node 11) and a previous key node is (namely, key node 1) meet matching detection condition, and current ordinary node (that is, ordinary node 11) and another previous key node (that is, key node 2) do not meet matching detection condition.This information processing method may be advanced to step S260, to judge current ordinary node (that is, ordinary node 11) and previous key node (that is, key node 1) whether Data Matching.

In step S260, judge current ordinary node and previous key node whether Data Matching.

When judging current ordinary node 11 and previously meeting described matching detection condition between key node 1, in order to carry out Data Matching detection, first, the integration observation data that previous key node 1 comprises can be read in.Then, can judge whether the frame observation data comprised at current ordinary node 11 mates with the integration observation data comprised at previous key node 1.If current ordinary node 11 and previously Data Matching between key node, then this information processing method may be advanced to step S270, to carry out figure optimizing process.Otherwise this information processing method turns back to step S210 to be continued to perform, to obtain a new frame observation data, until again cannot get a new frame observation data.

As illustrated in figure 5f, comprised the observation data in ordinary node 1-4 by the key node 1 represented at solid black circle, the subgraph M1 of local can be formed.When being filled current ordinary node 11 that circle represents by grey and finding closed loop, as long as carry out mate just passable with this key node 1.That is, when search loop connects, can only select key node as potential loop coupling target.The reason done like this is, different from the single frames observation data comprised at ordinary node, comprise a large amount of observation datas at Local Subgraphs, like this when such as carrying out observation data coupling by unique point, higher matching threshold can be set, to increase the accuracy of loop detection.

In one example, directly can determine that one about the conversion of pose between the two is initially guessed u based on the posture information of current ordinary node 11 and the posture information of previous key node 1.Such as, can move according to the code-disc of robot, find range model and Gaussian noise determine that this initially guesses u.

Next, this can be utilized initially to guess u, and use error equation, judge whether the frame observation data comprised at described current ordinary node 11 mates with the integration observation data comprised at described previous key node 1.

In the examples described above, a kind of mode for carrying out loop detection is provided.But when robot advances along very long path time, accumulative kinematic error becomes very large.In this case, this initially to guess that potential range can obtain the interval of convergence of globally optimal solution far.

Because the error equation related in frame matching algorithm is a kind of process asking locally optimal solution, if so when this initially guesses that u is positioned at the undesirable interval of convergence, even if there is very large similarity between the Current observation data comprised at current ordinary node 11 and the subgraph comprised at previous key node 1, this frame matching algorithm also may fall into the not good locally optimal solution of performance, and globally optimal solution cannot be obtained, thus the conversion that can not estimate exactly in loop coupling, final impact normally forms closed loop.

Fig. 6 illustrates the local search algorithm for closed loop detect according to the concrete example of the embodiment of the present invention.

As shown in Figure 6,4 intervals of convergence are comprised in local search algorithm.As shown in Figure 6, if initial guess is positioned at the position of u0, then this local search algorithm is only locally optimal solution s0 about the optimum solution that mode error finally obtains.For initial guess u1 and u3, this local search algorithm is also only locally optimal solution s1 and s3 about the optimum solution that mode error finally obtains.Only have when initial guess is positioned at the position of u2, this local search algorithm may be just globally optimal solution s2 about the optimum solution that mode error finally obtains.

In order to obtain globally optimal solution to form closed loop exactly, in another example, can by the multiple initial guess of interpolation adaptively between current ordinary node 11 and previous key node 1, make that there is an initially guess as far as possible in each interval of convergence, to guarantee to obtain a globally optimal solution in local search algorithm, thus accurately to detect closed circuit.

Fig. 7 illustrates and calculates ratio juris according to the loop detection based on interpolation method of the concrete example of the embodiment of the present invention.

As shown in Figure 7, such as, multiple pose (shown in triangle as little in grey) can be inserted linearly, to obtain multiple initial guess between the pose of current robot (shown in triangle as large in grey) and the potential key node pose (shown in triangle as large in black) that may form closed loop.

That is, multiple initial estimate can be generated according to the posture information of the posture information of current ordinary node 11 and previous key node 1.Such as, first, the spacing distance between described current ordinary node 11 and described previous key node 1 can be determined according to the posture information of current ordinary node 11 and the posture information of previous key node 1; Then, between a predetermined value and described spacing distance, select multiple value, as described multiple initial estimate.

Particularly, between predetermined value 0 and described spacing distance, N number of value can be inserted according to equal intervals, i.e. N number of pose,

Wherein, N is the number of the initial estimate be inserted into, be capping function, Dis is current ordinary node 11 and the spacing distance previously between key node 1, and T _loopthe convergence in mean being described error equation is interval, i.e. the mean value of interval of convergence 1-4 as shown in Figure 6.

Next, described multiple initial estimate can be utilized, and use error equation, judge whether the frame observation data comprised at described current ordinary node 11 mates with the integration observation data comprised at described previous key node 1.

Like this, by current ordinary node and the previously multiple initial guess of interpolation adaptively between key node, guarantee in local search algorithm, obtain a globally optimal solution, thus accurately detect closed circuit.

Preferably, in order to reduce the calculated amount of movable equipment or region server, can limit the number of initial guess.

In one example, can by considering that the error range of this present node limits the initial guess participating in frame matching algorithm.For this reason, when each ordinary node of generation, the error range of current ordinary node can be determined further.

Particularly, while such as scan matching method can be used in step S220 to carry out frame matching to this current ordinary node, variance and/or the covariance of current ordinary node is calculated by the operation result of this scan matching method.This variance and covariance can indicate the error range between the coordinate (x, y) of calculated current ordinary node in map and the true coordinate of this present node in circumstances not known.

For ordinary node 1, because (namely current robot is in initial position, position 1) place, so generally, often there is not any error in the coordinate information of this node 1 obtained, so after determining the error range of this node 1, probably learn that the error range of this node 1 is 0.For ordinary node 2, because now robot has left oneself initial position and the undefined position arrived in circumstances not known, so must certain error be there is in the coordinate position of this robot calculated by SLAM, and this error is also by the continuous motion along with robot, the continuous execution of calculating and adding up gradually.In other words, the error range of current ordinary node must be greater than the error range of first node above in history ordinary node, and the error range of first node also must be greater than the error range of second node above in history node above, and by that analogy.

After the error range obtaining current ordinary node 11, owing to can determine that initial guess must be within this error range, so the initial guess be in error range only can be chosen, to perform the follow-up loop detection algorithm for searching for globally optimal solution.

In step S270, the posture information of knot modification.

With the addition of the closed limit of loop between current ordinary node 11 and key node 1 after, can reduce by figure optimized algorithm the cumulative movement error that robot produces due to persistent movement.

The SLAM method realized based on this information processing method is with the SLAM method significant difference of the improvement described in the introduction, in this information processing method, any one ordinary node corresponding with it is not deleted after generating key node, therefore, in this step, the posture information of all ordinary nodes and key node can be adjusted according to the current posture information comprised at current ordinary node 11.

In a first example, each node in all ordinary nodes and key node can be selected as adjustment aim node, to carry out posture information adjustment.Such as, back-track algorithm can be adopted to perform posture information adjustment, that is, each node can be selected according to the backward of robot working line, as adjustment aim node.Namely, as illustrated in figure 5f, after closed circuit being detected, in turn can adjust the posture information of ordinary node 11, ordinary node 10, key node 3, ordinary node 8, ordinary node 7, ordinary node 6, key node 2, ordinary node 4, ordinary node 3, ordinary node 2, key node 1.

Such as, first can redefine the posture information of ordinary node 10 according to the current posture information comprised at current ordinary node 11, and use the posture information redefined to replace the original posture information of ordinary node 10.Then, can according to the current posture information comprised at current ordinary node 11 or the posture information redefining ordinary node 9 according to the posture information that ordinary node 10 after the adjustment comprises, and use the posture information redefined to replace the original posture information of ordinary node 9.The like, until the posture information of all nodes all adjusts complete.

In the examples described above, a kind of mode for carrying out pose adjustment is provided.But the present inventor finds, above-mentioned pose adjustment mode process there is high confidence level but the motion estimated by mistake or there is low confidence but the motion be estimated correctly time may there is deterioration.This deterioration especially runs through at mobile electronic equipment the place being similar to corridor and so on and easily occurs, in the place being similar to corridor and so on, frame alignment algorithm always produces high confidence level along corridor.But meanwhile, mobile electronic equipment occurs that skidding, accumulative telemetry difference or laser ranging method can cause high kinematic error.When this occurs, above-mentioned pose adjustment mode may lose efficacy, and this and figure optimize for maintaining " good " limit, improve the target on " bad " limit simultaneously is not obviously inconsistent.

Fig. 8 illustrates and adjusts failed scene according to the pose that may cause of the concrete example of the embodiment of the present invention.

As shown in Figure 8, when the scene by being similar to corridor, between current ordinary node 11 (shown in triangle as large in grey) and previous key node 1 (shown in triangle as large in black), produce closed circuit.In this graph structure, good limit (as shown in the solid arrow between node) be inserted in be similar to corridor scene outside, and bad limit (as shown in the dotted arrow between node) be inserted in be similar to corridor scene within, thus to estimate with being had deviation.If the result according to closed loop distributes pose adjusted value unitedly fifty-fifty or according to degree of confidence, then may cause originally estimating that good limit becomes deterioration.

In order to solve the problem, after optimization figure, the conversion on each limit can be recalculated.When upgrading the pose of robot, identical scan matching device method can be used to utilize multiple initial guess to calculate side information.For " good " limit, following frame alignment algorithm still can fall into the locally optimal solution of identical convergence, and this maintains the correct part of movement locus; And for " bad " limit, when one of initial guess better interval of convergence of hit, this algorithm may better be separated.

Therefore, in the second example, the multiple initial guess of interpolation adaptively between the posture information after the original posture information and this knot adjustment of each adjustment aim node (comprising ordinary node and key node) can be passed through, make that there is an initially guess as far as possible in each interval of convergence, to guarantee to obtain a globally optimal solution in local search algorithm, thus but but correctly process has high confidence level the motion estimated by mistake or have low confidence the motion be estimated correctly.

Fig. 9 illustrates the principle of the pose adjustment algorithm based on interpolation method according to the concrete example of the embodiment of the present invention.

As shown in Figure 9, such as, multiple pose (shown in triangle as little in grey) can be inserted linearly, to obtain multiple initial guess between pose (shown in triangle as large in black) after the original pose of adjustment aim node (shown in triangle as large in grey) and adjustment.

That is, multiple initial estimate can be generated according to the original pose of adjustment aim node with through the pose that the adjustment of the first example obtains.Such as, for key node 3 as adjustment aim node, can first, according to the spacing distance that the posture information after the original posture information of key node 3 and adjustment is determined between the two; Then, between a predetermined value and described spacing distance, select multiple value, as described multiple initial estimate.

Particularly, between predetermined value 0 and described spacing distance, N number of value can be inserted according to equal intervals, i.e. N number of pose,

Wherein, N is the number of the initial estimate be inserted into, be capping function, Dis is the spacing distance between the original pose of adjustment aim node and the pose after adjusting, and T _loopthe convergence in mean being described error equation is interval, i.e. the mean value of interval of convergence 1-4 as shown in Figure 6.

Next, described multiple initial estimate can be utilized, and use error equation, determine that is determined a final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum.Then, just can use determined final estimated value as object pose information to adjust the original posture information of described adjustment aim node, instead of the posture information after directly utilizing adjustment carries out pose adjustment.

Like this, by the multiple initial guess of interpolation adaptively, guarantee in local search algorithm, obtain a globally optimal solution, thus accurately frame matching is carried out in detection.

Preferably, in order to reduce the calculated amount of movable equipment or region server, can limit the number of initial guess.

In one example, can by considering that the error range of this adjustment aim node limits the initial guess participating in frame matching algorithm.For this reason, similarly, when each ordinary node of generation, the error range of current ordinary node can be determined further.

Particularly, while such as scan matching method can be used in step S220 to carry out frame matching to this current ordinary node, variance and/or the covariance of current ordinary node is calculated by the operation result of this scan matching method.This variance and covariance can indicate the error range between the coordinate (x, y) of calculated current ordinary node in map and the true coordinate of this present node in circumstances not known.

For ordinary node 1, because (namely current robot is in initial position, position 1) place, so generally, often there is not any error in the coordinate information of this node 1 obtained, so after determining the error range of this node 1, probably learn that the error range of this node 1 is 0.For ordinary node 2, because now robot has left oneself initial position and the undefined position arrived in circumstances not known, so must certain error be there is in the coordinate position of this robot calculated by SLAM, and this error is also by the continuous motion along with robot, the continuous execution of calculating and adding up gradually.In other words, the error range of current ordinary node must be greater than the error range of first node above in history ordinary node, and the error range of first node also must be greater than the error range of second node above in history node above, and by that analogy.

After the error range obtaining current ordinary node 11, owing to can determine that initial guess must be within this error range, so the initial guess be in error range only can be chosen, alternatively estimated value, only to utilize described candidate's estimated value, and use error equation, determine final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum, thus obtain a globally optimal solution.

It should be noted that, although for backward adjust each node posture information be illustrated, the present invention is not limited thereto.Such as, side by side can also adjust the posture information of each node or adjust the posture information of each node with the positive sequence of robot working line or other orders.

In step S280, reconstruct key node.

As mentioned above, because this information processing method is when carrying out pose adjustment to node, key node and ordinary node are identical treating, so can reconstruct the subgraph of key node according to the pose after the renewal of all ordinary nodes comprised at a key node.The motivation of reconstruct key node is, because the pose of the ordinary node comprised at key node adjusts, so reconstruct key node can be removed with the observation data that deviation merges in Local Subgraphs, to form higher level node more accurately.

Therefore, in this step, after figure optimization process, graph structure can be rebuild to keep the node with " good " limit as it is, and recalculates the node and limit with estimation of deviation, and upgrade the latter.

In one example, for each key node, such as, for key node 3, can according to all ordinary nodes after the adjustment comprised at this key node 3 (namely, ordinary node 9-11) posture information the multiframe observation data comprised at described all ordinary nodes is reintegrated, with obtain upgrade after integration observation data.Then, the integration observation data after described renewal can be used to replace original integration observation data, thus reconstruct described key node.Such as, simply, step S240 can be re-executed, to reintegrate the multiframe observation data comprised at these ordinary nodes according to the posture information after the renewal of ordinary node 9-11, to obtain integrating observation data, and use described integration observation data to replace the frame observation data comprised at ordinary node 9, thus this ordinary node 9 is converted to key node 3 again.

In another example, if key node formation condition triggers generation key node based on the range ability of robot or course length, so along with after the posture information that have adjusted node in step S270, the number of the ordinary node in a key node is probably caused also to occur change.At this moment, multiple ordinary node can be rejudged according to the posture information of all ordinary nodes after adjustment and whether meet key node formation condition, thus regenerate key node according to described multiple ordinary node.

In order to determine key node by which ordinary node stipulations is formed, can after generating key node based on selected ordinary node, in this key node, except the posture information of the integration observation data He this original ordinary node that comprise key node, the observation data of this original ordinary node of further reservation, to reconstruct key node.

But, in order to save storage space, the integration observation data of key node and the posture information of the original ordinary node corresponding with it only may be comprised in this key node.For this reason, according to the integration observation data comprised at this key node and posture information and the integration observation data comprised at the residue ordinary node forming this key node and posture information, this key node can be inversely transformed into ordinary node.

For key node 2, this key node 2 is the integration observation datas according to comprising at multiple ordinary node 5-8, generates based on the posture information of ordinary node 5.In order to this key node 2 is inversely transformed into ordinary node 5, remaining ordinary node can be used in subgraph M2 (namely, ordinary node 6-8) filter the integration observation data of multiple ordinary node 5-8 that this key node 2 comprises, to obtain a frame observation data of original ordinary node 5, and using the posture information after the adjustment of key node 2 directly as the posture information of this ordinary node 5, thus realize inverse transformation, for follow-up restructuring procedure.

In step S290, upgrade global map.

After regenerating whole subgraph M1-M3, can by re-starting frame matching between every two subgraphs, based on higher level node and regenerate global map.

As can be seen here, according in the information processing method of the concrete example of the embodiment of the present invention, propose a kind of figure based on key node to build and optimisation strategy, wherein, first, key node (key-node) comprises the information of subgraph (submap), may be used for more robust and accurately detects closed loop; Meanwhile, key node, when carrying out figure optimization, can be regarded as ordinary node, participates in figure optimization process, to reconstruct picture information after the optimization, thus when mistake appears in subgraph, can upgrade subgraph.

Below, with reference to figure 10A and Figure 10 B, above-mentioned effect is described intuitively.

Figure 10 A illustrates the design sketch of the map generated according to the SLAM method of prior art, and Figure 10 B illustrates the design sketch of the map generated according to the SLAM method of the embodiment of the present invention.

Illustrated in Figure 10 A, be made up of outer ring irregular curve according to the global map that the SLAM method of prior art generates, obviously, this irregular curve exists significantly different from the station figure in background, this is because this SLAM method causes a large amount of erroneous matching when generating map, so cause the final map generated to produce distortion.

But, illustrated in Figure 10 B, when using information processing method (SLAM method) according to the embodiment of the present invention, can find out that the station figure in the global map and background finally generated is very approximate, as shown in the irregular curve of outer ring.

In addition, according in the information processing method of the concrete example of the embodiment of the present invention, also proposed a kind of iterative optimisation strategy based on graph structure, wherein, first, linear difference can be carried out based on interval of convergence size (convergentpoolsize), estimate to obtain one group of initial pose, then call frame matching (scan-matcher) algorithm according to these initial estimation, thus obtain optimum loop coupling (loopmatch); Meanwhile, the error range at current such as edge variance (marginalcovariance) and so on can be considered, to prevent wrong closed loop (falseloop) and to increase closed-loop precision.

Below, with reference to figure 10C and Figure 10 D, above-mentioned effect is described intuitively.

Figure 10 C illustrates the design sketch of the map generated according to the SLAM method of prior art, and Figure 10 D illustrates the design sketch of the map generated according to the SLAM method of the embodiment of the present invention.

Illustrated in Figure 10 C, the trajectory diagram of the robot generated according to the SLAM method of prior art is made up of inner ring irregular curve, and obviously, this irregular curve fails to form closed circuit well.

But, illustrated in Figure 10 D, when using information processing method (SLAM method) according to the embodiment of the present invention, can find out that the trajectory diagram of the final robot generated defines closed circuit well.

In addition, according in the information processing method of the concrete example of the embodiment of the present invention, also proposed another iterative optimisation strategy based on graph structure, wherein, first, linear difference can be carried out based on interval of convergence size (convergentpoolsize), estimate to obtain one group of initial pose, then call frame matching (scan-matcher) algorithm according to these initial estimation, thus obtain optimum estimation (motionestimation); Meanwhile, the error range at current such as edge variance (marginalcovariance) and so on can be considered, to prevent Outlier match (outliermatches) and to increase the coupling alignment precision between node.

Below, with reference to figure 10E and Figure 10 F, above-mentioned effect is described intuitively.

Figure 10 E illustrates the design sketch of the map generated according to the SLAM method of prior art, and Figure 10 F illustrates the design sketch of the map generated according to the SLAM method of the embodiment of the present invention.

Illustrated in Figure 10 E, be made up of outer ring irregular curve according to the global map that the SLAM method of prior art generates, obviously, this irregular curve exists significantly different from the station figure in background, this is because this SLAM method causes a large amount of Outlier matchs when carrying out figure and optimizing, so cause the final map generated to produce distortion.

But, illustrated in Figure 10 F, when using information processing method (SLAM method) according to the embodiment of the present invention, can find out that the station figure in the global map and background finally generated is very approximate, as shown in the irregular curve of outer ring.

It should be noted that, although describe the concrete example according to the embodiment of the present invention according to the pre-arranged procedure order above, the present invention is not limited thereto.But, above-mentioned information processing method can be realized according to other different sequence of steps.

Such as, in this information processing method, step S230-S240 performs after can being placed on step S250-S260, namely when judging current ordinary node and previously do not had Data Matching between key node in step S260, can perform step S230.

In addition, it should be noted that, although by the posture information adjustment mode based on interpolation (namely, step S270) be used in based on key node SLAM method in be described, but, obviously, SLAM method only based on the SLAM method of ordinary node or the improvement based on ordinary node and super node can should be applied to equally based on posture information adjustment mode of interpolation.

Figure 11 illustrates according to signal conditioning package of the present invention, and Figure 12 illustrates the mobile electronic equipment according to the embodiment of the present invention.

The information processing method according to the embodiment of the present invention illustrated in Fig. 1 and Fig. 2 can be realized by the signal conditioning package 100 illustrated in Figure 11, and this signal conditioning package 100 can be applied to the one or more mobile electronic equipments 1000 illustrated in Figure 12.

In one embodiment, described mobile electronic equipment 1000 can be robot (Robot), for moving in targeted environment, simultaneously by performing instant location with map structuring (SLAM) to self carrying out quick position and generating the diagram data exactly about this targeted environment.

In one example, described mobile electronic equipment 1000 can independently move in targeted environment, performs instant location and map structuring (SLAM) simultaneously.

In another example, described mobile electronic equipment 1000 also can with other electronic equipments (such as, region server) communicate, move in targeted environment with the map datum provided according to region server and Route Planning Data, perform instant location and map structuring (SLAM) simultaneously, and send the result of described instant location and map structuring to region server, for upgrading the map datum stored in described region server.

As illustrated in Figure 12, this mobile electronic equipment 1000 can comprise: signal conditioning package 100, data collector 200.

This signal conditioning package 100 may be used for while generating key node according at least one ordinary node, retain this at least one ordinary node, and when judging that the frame observation data comprised at current ordinary node is mated with the integration observation data comprised at previous key node, the posture information of all nodes is adjusted according to the current posture information comprised at described current ordinary node, all key nodes are reconstructed according to the posture information of all ordinary nodes after adjustment and the posture information of all key nodes, thus upgrade global map according to all key nodes after reconstruct.

This image collecting device 200 may be used for gathering observation data.Such as, this image collecting device 200 can be laser sensor, it is first-class to make a video recording, and is respectively used to obtain laser scanning result, image observation result or disparity map etc.

In addition, when needing to communicate with other electronic equipments of such as region server and so on, this mobile electronic equipment 1000 can also comprise: communicator 300.

This communicator 300 may be used for communicating with other mobile electronic equipments or server, to receive or to send observation data, path planning, local map, global map etc. to other mobile electronic equipments.

In addition, this signal conditioning package 100 can be communicated with mobile electronic equipment 1000 by any mode.

In one example, this signal conditioning package 100 can be integrated in this mobile electronic equipment 1000 as a software module and/or hardware module, and in other words, this mobile electronic equipment 1000 can comprise this signal conditioning package 100.Such as, when mobile electronic equipment 1000 is robots, this signal conditioning package 100 can be a software module in the operating system of this robot, or can be aimed at the application program that this robot develops; Certainly, this signal conditioning package 100 can be one of numerous hardware modules of this robot equally.

Alternatively, in another example, this signal conditioning package 100 and this mobile electronic equipment 1000 also can be the equipment be separated, and this signal conditioning package 100 can be connected to this mobile electronic equipment 1000 by wired and/or wireless network, and transmit interactive information according to the data layout of agreement.

As illustrated in Figure 11, can comprise according to the signal conditioning package 100 of the embodiment of the present invention: ordinary node generation unit 110, Data Matching judging unit 120, posture information adjustment unit 130, key node reconfiguration unit 140 and global map updating block 150.

This ordinary node generation unit 110 may be used for generating current ordinary node according to a frame observation data of current acquisition and the current posture information of described mobile electronic equipment.

This Data Matching judging unit 120 may be used for judging whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at previous key node, wherein, described integration observation data is by integrating obtained to the multiframe observation data comprised at multiple ordinary node.

This posture information adjustment unit 130 may be used for, when judging that the frame observation data comprised at described current ordinary node is mated with the integration observation data comprised at described previous key node, adjusting the posture information of all nodes according to the current posture information comprised at described current ordinary node.

The posture information that this key node reconfiguration unit 140 may be used for all nodes after according to adjustment reconstructs all key nodes.

This global map updating block 150 may be used for all key nodes after according to reconstruct and upgrades global map.

In one embodiment, described device can also comprise: key node converting unit 160.

This key node converting unit 160 may be used for after the current ordinary node of generation, when the multiple ordinary nodes judging to comprise described current ordinary node meet key node formation condition, from the multiple ordinary nodes comprising described current ordinary node, select an ordinary node; Posture information according to described multiple ordinary node is integrated the multiframe observation data comprised at described multiple ordinary node, to obtain integrating observation data; And use described integration observation data to replace the frame observation data comprised at selected ordinary node, thus selected ordinary node is converted to current key node.

In one embodiment, described device also comprises: matching condition judging unit 170 and key node reading unit 180.

This matching condition judging unit 170 may be used for, before whether the frame observation data judging to comprise at described current ordinary node mates with the integration observation data comprised at previous key node, judging whether meet matching detection condition between described current ordinary node and described previous key node.

This key node reading unit 180 may be used for, when judging to meet described matching detection condition between described current ordinary node and described previous key node, reading in the integration observation data that described previous key node comprises.

In one example, whether described matching condition judging unit 170 can meet matching detection condition by current ordinary node described in following operation judges and between described previous key node: calculate the spacing distance between described current ordinary node and described previous key node according to the posture information of described current ordinary node and the posture information of described previous key node; More described spacing distance and predetermined threshold; When described spacing distance is less than or equal to described predetermined threshold, judge to meet described matching detection condition between described current ordinary node and described previous key node; And when described spacing distance is greater than described predetermined threshold, judge not meet described matching detection condition between described current ordinary node and described previous key node.

In one embodiment, whether the frame observation data that described Data Matching judging unit 120 can be comprised at described current ordinary node by following operation judges mates with the integration observation data comprised at previous key node: generate multiple initial estimate according to the posture information of described current ordinary node and the posture information of described previous key node; And utilize described multiple initial estimate, and use error equation, judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at described previous key node.

In one example, described Data Matching judging unit 120 can generate multiple initial estimate by following operation according to the posture information of the posture information of described current ordinary node and described previous key node: determine the spacing distance between described current ordinary node and described previous key node according to the posture information of described current ordinary node and the posture information of described previous key node; And between predetermined value and described spacing distance, select multiple value, as described multiple initial estimate.

In one example, described Data Matching judging unit 120 can select multiple value to comprise by following operating between predetermined value and described spacing distance: between described predetermined value and described spacing distance, select N number of value according to equal intervals,

Wherein, N is the number of initial estimate, be capping function, Dis is described spacing distance, and T _loopthe convergence in mean being described error equation is interval.

In one example, described Data Matching judging unit 120 can utilize described multiple initial estimate by following operation, and use error equation, judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at described previous key node: the error range determining described current ordinary node, described error range indicates the error between posture information and the attained pose of described mobile electronic equipment comprised at described current ordinary node; The initial estimate be in described error range is selected, alternatively estimated value among described multiple initial estimate; And only utilize described candidate's estimated value, and use error equation, judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at described previous key node.

In one embodiment, described posture information adjustment unit 130 can adjust the posture information of all nodes according to the current posture information comprised at described current ordinary node by following operation: select each node as adjustment aim node; The posture information of described adjustment aim node is redefined according to the current posture information comprised at described current ordinary node; And the original posture information of described adjustment aim node is at least adjusted according to the posture information redefined.

In one example, described posture information adjustment unit 130 at least can adjust the original posture information of described adjustment aim node according to the posture information redefined by following operation: the original posture information according to described adjustment aim node generates multiple initial estimate with the posture information redefined; Utilize described multiple initial estimate, and use error equation, determine final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum; And use described final estimated value as object pose information to adjust the original posture information of described adjustment aim node.

In one example, described posture information adjustment unit 130 can utilize described multiple initial estimate by following operation, and use error equation, determine final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum: the error range determining described adjustment aim node, described error range indicates the error between posture information and the attained pose of described mobile electronic equipment comprised at described adjustment aim node; The initial estimate be in described error range is selected, alternatively estimated value among described multiple initial estimate; And only utilize described candidate's estimated value, and use error equation, determine final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum.

In one embodiment, described key node reconfiguration unit 140 can reconstruct all key nodes by following operation according to the posture information of all nodes after adjustment: for each key node, posture information according to all ordinary nodes after the adjustment comprised at described key node is reintegrated the multiframe observation data comprised at described all ordinary nodes, to obtain the integration observation data after upgrading; And use the integration observation data after described renewal to replace original integration observation data, thus reconstruct described key node.

Concrete configuration according to the unit in the signal conditioning package 100 of the embodiment of the present invention and each device in mobile electronic equipment 1000 is introduced in detail with operation in the information processing method described above with reference to Fig. 1 and 2, and therefore, its repeated description will be omitted.

As can be seen here, adopt the signal conditioning package according to the embodiment of the present invention, can while generating key node according at least one ordinary node, retain this at least one ordinary node, and when judging that the frame observation data comprised at current ordinary node is mated with the integration observation data comprised at previous key node, the posture information of all nodes is adjusted according to the current posture information comprised at described current ordinary node, all key nodes are reconstructed according to the posture information of all ordinary nodes after adjustment and the posture information of all key nodes, thus upgrade global map according to all key nodes after reconstruct.Therefore, in an embodiment of the present invention, because key node comprises abundant picture information, thus can more robust and accurately detection loop close; And ordinary node can be regarded as when figure optimizes due to key node and participate in optimizing to reconstruct picture information, so when carrying out map in current in higher level, eliminate the inconsistent frame matching existed in meromixis observation data well.

In addition, although above-mentioned unit is illustrated each embodiment of the present invention as the executive agent of each step herein, those skilled in the art are it is understood that the present invention is not limited thereto.The executive agent of each step can be served as by other one or more units, unit, even module.

Such as, each step performed by above-mentioned receiving element 110, resolution unit 120, determining unit 130 and applying unit 140 can be realized by the CPU (central processing unit) (CPU) in mobile electronic equipment uniformly.

Through the above description of the embodiments, those skilled in the art can be well understood to the mode that the present invention can add required hardware platform by means of software and realize, and can certainly all be implemented by software or hardware.Based on such understanding, what technical scheme of the present invention contributed to background technology can embody with the form of software product in whole or in part, this computer software product can be stored in storage medium, as ROM/RAM, disk, CD etc., comprising some instructions in order to make a computer equipment (can be personal computer, server, or the network equipment etc.) perform the method described in some part of each embodiment of the present invention or embodiment.

Each embodiment of the present invention is described in detail above.But, it should be appreciated by those skilled in the art that without departing from the principles and spirit of the present invention, various amendment can be carried out to these embodiments, combination or sub-portfolio, and such amendment should fall within the scope of the present invention.

Claims (24)

1. an information processing method, described method is applied to mobile electronic equipment, and described mobile electronic equipment is used to carry out map structuring, it is characterized in that, described method comprises:

Current ordinary node is generated according to a frame observation data of current acquisition and the current posture information of described mobile electronic equipment;

Judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at previous key node, wherein, described integration observation data is by integrating obtained to the multiframe observation data comprised at multiple ordinary node;

When judging that the frame observation data comprised at described current ordinary node is mated with the integration observation data comprised at described previous key node, adjust the posture information of all nodes according to the current posture information comprised at described current ordinary node;

Posture information according to all nodes after adjustment reconstructs all key nodes; And

Global map is upgraded according to all key nodes after reconstruct.

2. method according to claim 1, is characterized in that, after the current ordinary node of generation, described method also comprises:

When the multiple ordinary nodes judging to comprise described current ordinary node meet key node formation condition, from the multiple ordinary nodes comprising described current ordinary node, select an ordinary node;

Posture information according to described multiple ordinary node is integrated the multiframe observation data comprised at described multiple ordinary node, to obtain integrating observation data; And

Use described integration observation data to replace the frame observation data comprised at selected ordinary node, thus selected ordinary node is converted to current key node.

3. method according to claim 1, is characterized in that, before whether the frame observation data judging to comprise at described current ordinary node mates with the integration observation data comprised at previous key node, described method also comprises:

Judge whether meet matching detection condition between described current ordinary node and described previous key node; And

When judging to meet described matching detection condition between described current ordinary node and described previous key node, read in the integration observation data that described previous key node comprises.

4. method according to claim 3, is characterized in that, judges that whether meeting matching detection condition between described current ordinary node and described previous key node comprises:

The spacing distance between described current ordinary node and described previous key node is calculated according to the posture information of described current ordinary node and the posture information of described previous key node;

More described spacing distance and predetermined threshold;

When described spacing distance is less than or equal to described predetermined threshold, judge to meet described matching detection condition between described current ordinary node and described previous key node; And

When described spacing distance is greater than described predetermined threshold, judge not meet described matching detection condition between described current ordinary node and described previous key node.

5. method according to claim 1, is characterized in that, judges whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at previous key node and comprises:

Multiple initial estimate is generated according to the posture information of described current ordinary node and the posture information of described previous key node; And

Utilize described multiple initial estimate, and use error equation, judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at described previous key node.

6. method according to claim 5, is characterized in that, generates multiple initial estimate comprise according to the posture information of described current ordinary node and the posture information of described previous key node:

The spacing distance between described current ordinary node and described previous key node is determined according to the posture information of described current ordinary node and the posture information of described previous key node; And

Multiple value is selected, as described multiple initial estimate between predetermined value and described spacing distance.

7. method according to claim 6, is characterized in that, selects multiple value to comprise between predetermined value and described spacing distance:

N number of value is selected according to equal intervals between described predetermined value and described spacing distance,

Wherein, N is the number of initial estimate, be capping function, Dis is described spacing distance, and T _loopthe convergence in mean being described error equation is interval.

8. method according to claim 5, it is characterized in that, utilize described multiple initial estimate, and use error equation, judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at described previous key node and comprise:

Determine the error range of described current ordinary node, described error range indicates the error between posture information and the attained pose of described mobile electronic equipment comprised at described current ordinary node;

The initial estimate be in described error range is selected, alternatively estimated value among described multiple initial estimate; And

Only utilize described candidate's estimated value, and use error equation, judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at described previous key node.

9. method according to claim 1, is characterized in that, the posture information adjusting all nodes according to the current posture information comprised at described current ordinary node comprises:

Select each node as adjustment aim node;

The posture information of described adjustment aim node is redefined according to the current posture information comprised at described current ordinary node; And

The original posture information of described adjustment aim node is at least adjusted according to the posture information redefined.

10. method according to claim 9, is characterized in that, the original pose packets of information that the posture information that at least basis redefines adjusts described adjustment aim node is drawn together:

Original posture information according to described adjustment aim node generates multiple initial estimate with the posture information redefined;

Utilize described multiple initial estimate, and use error equation, determine final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum; And

Use described final estimated value as object pose information to adjust the original posture information of described adjustment aim node.

11. methods according to claim 10, is characterized in that, utilize described multiple initial estimate, and use error equation, determine final estimated value, to make, the matching error between described adjustment aim node and adjacent node is minimum to be comprised:

Determine the error range of described adjustment aim node, described error range indicates the error between posture information and the attained pose of described mobile electronic equipment comprised at described adjustment aim node;

The initial estimate be in described error range is selected, alternatively estimated value among described multiple initial estimate; And

Only utilize described candidate's estimated value, and use error equation, determine final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum.

12. methods according to claim 1, is characterized in that, reconstruct all key nodes comprise according to the posture information of all nodes after adjustment:

For each key node,

Posture information according to all ordinary nodes after the adjustment comprised at described key node is reintegrated the multiframe observation data comprised at described all ordinary nodes, to obtain the integration observation data after upgrading; And

Use the integration observation data after described renewal to replace original integration observation data, thus reconstruct described key node.

13. 1 kinds of signal conditioning packages, described application of installation is in mobile electronic equipment, and described mobile electronic equipment is used to carry out map structuring, it is characterized in that, described device comprises:

Ordinary node generation unit, for generating current ordinary node according to a frame observation data of current acquisition and the current posture information of described mobile electronic equipment;

Data Matching judging unit, for judging whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at previous key node, wherein, described integration observation data is by integrating obtained to the multiframe observation data comprised at multiple ordinary node;

Posture information adjustment unit, for when judging that the frame observation data comprised at described current ordinary node is mated with the integration observation data comprised at described previous key node, adjust the posture information of all nodes according to the current posture information comprised at described current ordinary node;

Key node reconfiguration unit, for reconstructing all key nodes according to the posture information of all nodes after adjustment; And

Global map updating block, for upgrading global map according to all key nodes after reconstruct.

14. devices according to claim 13, is characterized in that, described device also comprises:

Key node converting unit, for after the current ordinary node of generation, when the multiple ordinary nodes judging to comprise described current ordinary node meet key node formation condition, from the multiple ordinary nodes comprising described current ordinary node, select an ordinary node; Posture information according to described multiple ordinary node is integrated the multiframe observation data comprised at described multiple ordinary node, to obtain integrating observation data; And use described integration observation data to replace the frame observation data comprised at selected ordinary node, thus selected ordinary node is converted to current key node.

15. devices according to claim 13, is characterized in that, described device also comprises:

Matching condition judging unit, before whether mating with the integration observation data comprised at previous key node in the frame observation data judging to comprise at described current ordinary node, judge whether meet matching detection condition between described current ordinary node and described previous key node; And

Key node reading unit, for when judging to meet described matching detection condition between described current ordinary node and described previous key node, reads in the integration observation data that described previous key node comprises.

16. devices according to claim 15, is characterized in that, whether described matching condition judging unit meets matching detection condition by current ordinary node described in following operation judges and between described previous key node:

The spacing distance between described current ordinary node and described previous key node is calculated according to the posture information of described current ordinary node and the posture information of described previous key node;

More described spacing distance and predetermined threshold;

When described spacing distance is less than or equal to described predetermined threshold, judge to meet described matching detection condition between described current ordinary node and described previous key node; And

When described spacing distance is greater than described predetermined threshold, judge not meet described matching detection condition between described current ordinary node and described previous key node.

17. devices according to claim 13, is characterized in that, whether the frame observation data that described Data Matching judging unit is comprised at described current ordinary node by following operation judges mates with the integration observation data comprised at previous key node:

Multiple initial estimate is generated according to the posture information of described current ordinary node and the posture information of described previous key node; And

Utilize described multiple initial estimate, and use error equation, judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at described previous key node.

18. devices according to claim 17, is characterized in that, described Data Matching judging unit generates multiple initial estimate by following operation according to the posture information of the posture information of described current ordinary node and described previous key node:

The spacing distance between described current ordinary node and described previous key node is determined according to the posture information of described current ordinary node and the posture information of described previous key node; And

Multiple value is selected, as described multiple initial estimate between predetermined value and described spacing distance.

19. devices according to claim 18, is characterized in that, described Data Matching judging unit selects multiple value to comprise by following operating between predetermined value and described spacing distance:

N number of value is selected according to equal intervals between described predetermined value and described spacing distance,

Wherein, N is the number of initial estimate, be capping function, Dis is described spacing distance, and T _loopthe convergence in mean being described error equation is interval.

20. devices according to claim 17, it is characterized in that, described Data Matching judging unit utilizes described multiple initial estimate by following operation, and use error equation, judges whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at described previous key node:

Determine the error range of described current ordinary node, described error range indicates the error between posture information and the attained pose of described mobile electronic equipment comprised at described current ordinary node;

The initial estimate be in described error range is selected, alternatively estimated value among described multiple initial estimate; And

Only utilize described candidate's estimated value, and use error equation, judge whether the frame observation data comprised at described current ordinary node mates with the integration observation data comprised at described previous key node.

21. devices according to claim 13, is characterized in that, described posture information adjustment unit adjusts the posture information of all nodes according to the current posture information comprised at described current ordinary node by following operation:

Select each node as adjustment aim node;

The posture information of described adjustment aim node is redefined according to the current posture information comprised at described current ordinary node; And

The original posture information of described adjustment aim node is at least adjusted according to the posture information redefined.

22. devices according to claim 21, is characterized in that, described posture information adjustment unit at least adjusts the original posture information of described adjustment aim node according to the posture information redefined by following operation:

Original posture information according to described adjustment aim node generates multiple initial estimate with the posture information redefined;

Utilize described multiple initial estimate, and use error equation, determine final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum; And

Use described final estimated value as object pose information to adjust the original posture information of described adjustment aim node.

23. devices according to claim 22, it is characterized in that, described posture information adjustment unit utilizes described multiple initial estimate by following operation, and use error equation, determine final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum:

Determine the error range of described adjustment aim node, described error range indicates the error between posture information and the attained pose of described mobile electronic equipment comprised at described adjustment aim node;

The initial estimate be in described error range is selected, alternatively estimated value among described multiple initial estimate; And

Only utilize described candidate's estimated value, and use error equation, determine final estimated value, to make the matching error between described adjustment aim node and adjacent node minimum.

24. devices according to claim 13, is characterized in that, described key node reconfiguration unit reconstructs all key nodes by following operation according to the posture information of all nodes after adjustment:

For each key node,

Posture information according to all ordinary nodes after the adjustment comprised at described key node is reintegrated the multiframe observation data comprised at described all ordinary nodes, to obtain the integration observation data after upgrading; And

Use the integration observation data after described renewal to replace original integration observation data, thus reconstruct described key node.