CN117723076A

CN117723076A - Underground mine truck unmanned path planning method and system based on improved Dijkstra

Info

Publication number: CN117723076A
Application number: CN202311529567.3A
Authority: CN
Inventors: 尹业华; 李云财; 巫乔顺; 李平; 皮坤
Original assignee: Yunnan Kungang Electronic Information Technology Co ltd
Current assignee: Yunnan Kungang Electronic Information Technology Co ltd
Priority date: 2023-11-16
Filing date: 2023-11-16
Publication date: 2024-03-19

Abstract

The invention relates to an underground mine truck unmanned path planning method and system based on improved Dijkstra, and belongs to the technical field of unmanned operation. The invention can improve the operation efficiency of the mine card and can provide higher safety in complex mine environment. By the method, mine enterprises can be helped to better cope with challenging environmental conditions, production efficiency is improved, operation cost is reduced, and overall operation safety is improved.

Description

Underground mine truck unmanned path planning method and system based on improved Dijkstra

Technical Field

The invention belongs to the technical field of unmanned, and particularly relates to an underground mine truck unmanned path planning method and system based on improved Dijkstra.

Background

Unmanned navigation of underground mine cards in an underground mine environment is a key task. Due to the complexity and challenges of the mine environment, efficient and safe path planning methods are indispensable. For example, the conventional Dijkstra algorithm suffers from a number of problems in dealing with such environments, including the inability to deal with dynamically changing mine environments, the inability to ensure true safety, and the lack of understanding of actual mine card movement characteristics. Thus, there is a need for a more advanced and practical approach to addressing these issues.

Disclosure of Invention

The core of the algorithm of the invention is the optimization of the traditional Dijkstra algorithm and the combination of a new mining card kinematics model. The kinematic model not only can simulate the real motion characteristics of the mine card, but also can effectively plan a path in a complex mine environment. By means of the novel hinged mining card kinematic model, the motion state of the mining card and problems possibly encountered can be predicted more accurately. For example, obstructions, terrain variations, and even the presence of other mining cards that may be encountered during travel are considered. These factors are difficult to handle for the traditional Dijkstra algorithm, but are easy for our new model. Furthermore, by means of the improved Dijkstra algorithm we will be able to better utilize and optimize the run time of the mine cards. This means that our algorithm can complete more shoveling tasks in the same time, or that our algorithm can save a lot of time when the same number of shoveling tasks are completed. At the same time, this also means that our algorithm can more effectively use the resources of the mine card, thus reducing the operating costs. More importantly, by combining a new articulated mining card kinematic model and a modified Dijkstra algorithm, we will be able to ensure the security of the unmanned mining card. The safety is not only in preventing collision between mine cards, but also in adaptability to complex mine environments and coping capability to sudden events. For example, when an emergency is encountered, our algorithm can quickly adjust the path of the mine card, avoiding possible collisions, thus minimizing possible losses.

The technical scheme of the invention is as follows:

an improved Dijkstra underground mine truck unmanned path planning method comprises the following steps:

step (1) data collection

Sensor data collection: performing regional scanning by using a laser radar to obtain well wall surface point cloud data; acquiring position and posture data of the unmanned vehicle by using an inertial navigation system; acquiring cargo box load information by using a load sensor; acquiring container position information by using a mechanical arm position sensor;

step (2) Path planning

Processing the point cloud data into a grid map, and marking the positions of the obstacles; marking a starting point, an ending point and a charging station position on a map; based on an improved Dijkstra algorithm, carrying out shortest path planning on a map; the path planning result comprises a global path route and local obstacle avoidance information.

Further, processing the point cloud data into node and side information, using key points in the feature extraction environment as nodes, and calculating the distance and direction between the nodes as side information by using the node data; as unmanned mining trucks move in the environment, the newly acquired sensor data is continually used to optimize and enhance the accuracy of the map; adopting a graph optimization algorithm to continuously adjust the node positions and delete the error nodes and edges so that the map is closest to the real environment; when the unmanned mining truck returns to the previously traversed place, loop detection is realized through feature matching, so that the accuracy of the map is further improved, and the node positioning is optimized.

Further, in the modified Dijkstra algorithm, the weight function is as follows:

w (u, v) =α TerrainWeight (u, v) +β geognosis weight (u, v) +γ langerweight t (u, v) +δ CostWeight (u, v); wherein:

α, β, γ, δ are weight coefficients for balancing different factors;

TerrainWeight (u, v) represents terrain weight, and is specified according to the type of terrain;

geologicweight (u, v) represents geological weights, and is specified according to geological conditions; dangerweight (u, v) represents a risk weight, manually assigning a weight of 10-20% to the edge considered to be dangerous;

CostWeight (u, v) represents the running cost weight, and the time or energy consumption cost under different environments is learned through a machine learning algorithm.

Further, in the modified Dijkstra algorithm, the constraint conditions are as follows:

avoiding dangerous area constraints: identifying dangerous areas existing underground in advance, and enabling a path to bypass the dangerous areas in the path searching process;

flatness constraint: preferably selecting a path with a smoother ground to ensure the stability of the unmanned vehicle;

turning radius constraint: the turning radius of the path cannot be smaller than the minimum turning radius of the unmanned vehicle;

vehicle body size constraints: the path width must be greater than the body size of the unmanned vehicle;

visibility constraints: preferably selecting a path with good illumination condition and wide visual field;

ground constraint: the vehicle body is always kept on the ground, so that suspension is avoided;

carrying weight constraints: the path must be able to withstand the load requirements of the unmanned vehicle;

mineral wall and pillar restraint: maintaining a safe distance from the mine wall and the support post.

Further, the improved path planning algorithm framework includes:

initializing: initializing a starting point, adding the starting point into a priority queue, and setting the initial path weight to be 0; initializing the path weight of each node to infinity, which indicates that the node has not been accessed; initializing an empty path set;

main cycle: the node with the minimum current path weight is taken out from the priority queue; for each neighbor of a node: calculating an original weight comprising distance or time; checking whether all the additional constraint conditions are met, and if so, calculating the weight of the additional constraint term; calculating the total weight, namely the weight of the original weight plus the additional constraint term; if the new path weight is less than the currently known path weight, updating the path weight and updating the path; if the node is not in the path set, adding the node into a priority queue;

termination condition: stopping the algorithm when the target node is added to the path set or the priority queue is empty;

and (3) path backtracking: backtracking from the end point to the start point to obtain an optimal path;

algorithm modification points: in calculating the original weight, the weight of the edge in the graph is used, including distance or time; for each additional constraint condition, modifying the weight function according to the problem requirement, and carrying out corresponding check in the algorithm; sorting in the priority queue by using the path weight of the node; in the case of dynamic weight adjustment, the weights are updated periodically or according to real-time data.

The invention also relates to an underground mine truck unmanned driving path planning system, which is characterized in that: the system comprises a laser radar, an inertial navigation system, a load sensor, a mechanical arm position sensor and a processor connected with the sensors, wherein the processor is also connected with a display and a memory, and the laser radar performs regional scanning to acquire wall surface point cloud data of a well; the inertial navigation system acquires position and posture data of the unmanned vehicle; the load sensor acquires cargo information of a cargo box; the mechanical arm position sensor acquires container position information; and the processor performs mining card unmanned driving path planning based on the acquired data, and the mining card unmanned driving path planning is performed according to the method.

The invention also relates to an electronic device comprising a memory, a processor and a computer program on the memory and executable on the processor, which processor implements the steps of the above method when executing the computer program.

The invention also relates to a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method as described above.

The invention develops an improved Dijkstra algorithm suitable for underground mines, and combines a pushed novel hinged mining card kinematic model to improve the path planning efficiency and unmanned safety. The algorithm not only can improve the operation efficiency of the mine card, but also can provide higher safety in complex mine environments. By the method, mine enterprises can be helped to better cope with challenging environmental conditions, production efficiency is improved, operation cost is reduced, and overall operation safety is improved.

Drawings

FIG. 1 is a schematic diagram of a model turn of an embodiment of the present invention;

FIG. 2 is a dimensional view of a mining truck model in accordance with an embodiment of the present invention; units: mm;

fig. 3 is a system block diagram of an embodiment of the present invention.

Detailed Description

The following description of the embodiments will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all, of the embodiments of the present application. Based on the embodiments, all other embodiments that may be made by one of ordinary skill in the art without making any inventive effort are within the scope of the present application.

Unless otherwise defined, technical or scientific terms used in the embodiments of the present application should be given a general meaning as understood by one of ordinary skill in the art. The terms "first," "second," and the like, as used in this embodiment, do not denote any order, quantity, or importance, but rather are used to distinguish one element from another. The word "comprising" or "comprises", and the like, means that elements or items preceding the word are included in the element or item listed after the word and equivalents thereof, but does not exclude other elements or items. "mounted," "connected," and "connected" are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. "upper", "lower", "left", "right", "transverse", and "vertical", etc. are used only with respect to the orientation of the components in the drawings, these directional terms are relative terms, which are used for descriptive and clarity with respect thereto and which may vary accordingly with respect to the orientation in which the components are disposed in the drawings.

The underground mine truck unmanned path planning method based on the improved Dijkstra of the embodiment comprises the following steps:

and (3) data collection in the step (1):

sensor data collection: performing regional scanning by using a laser radar to obtain well wall surface point cloud data; acquiring position and posture data of the unmanned vehicle by using an inertial navigation system; acquiring cargo box load information by using a load sensor; and acquiring container position information by using a mechanical arm position sensor.

Step (2) Path planning

Processing the point cloud data into a grid map, and marking the positions of the obstacles; marking a starting point, an ending point and a charging station position on a map; based on the improved Dijkstra algorithm, carrying out shortest path planning on a map; the path planning result comprises a global path route and local obstacle avoidance information.

Based on the improved Dijkstra algorithm, the method specifically comprises the following steps:

1. map modeling

Modeling the mine environment, and establishing a map in the underground mine environment mainly comprises the following steps: 1.1. defining a node: each milestone location (e.g., intersection, key location, etc.) is defined as a node in the map.

1.2. Defining an edge: the connection relationship between nodes is defined as an edge. The path from node a to node B is defined as an edge.

1.3. Acquiring map data: environmental data is acquired using a sensor such as a lidar. And acquiring point cloud data of environments such as mine channels, rooms and the like through continuous scanning.

1.4. And (3) map construction: and processing the point cloud data into node and side information. Key points in the environment may be extracted as nodes using feature extraction techniques. The distance and direction between the nodes are calculated as side information using the node data. The distance between nodes is calculated by using the time difference of the laser radar, and the direction between the nodes is based on the relative angle or direction vector between radar data points.

1.5. Graph optimization: newly acquired sensor data is continually used to optimize and enhance the accuracy of the map as unmanned mining trucks move around the environment. And continuously adjusting the node positions, deleting the error nodes and edges and the like by adopting a least square method in a graph optimization algorithm, so that the map is closest to the real environment.

1.6. And loop detection: when the unmanned mining truck returns to the previously traversed place, loop detection can be realized through NCC template matching in feature matching, so that the accuracy of the map is further improved and the node positioning is optimized.

The data are collected through the sensors, the characteristics are extracted, the nodes and edges are established, and the map is continuously optimized by utilizing technologies such as map optimization, loop detection and the like, so that a map accurately expressing the mine environment topological structure is finally established.

2. Improved weight function:

and modifying the weight function of the edge, and considering factors such as terrain, geology, regional danger and the like to ensure that the path planning reflects the complexity of the actual environment. Modifying the edge weight function according to the actual environmental factors:

2.1. additional edge attributes are defined to represent terrain, geology (hard rock, soft soil, etc.). Different weights may be assigned to different terrains, geology.

2.2. Defining the risk attribute manually assigns a large weight to the sides considered dangerous so that the path planning algorithm tends to avoid these sides.

2.3. The actual driving data are collected, and the time or energy consumption cost under different environments is learned through a linear regression algorithm in a machine learning algorithm to be used as the weight of the edge.

2.4. For terrain data, it is considered to map elevation changes to weights as well, with the mountain roads being weighted more heavily.

2.5. And modeling various environmental factors in comprehensive consideration, and performing multi-objective optimization, wherein the distance is considered, and indexes such as running cost, safety and the like are also considered.

2.6. When an accurate 3D map exists, path planning can be directly performed in a three-dimensional environment instead of simplifying the three-dimensional environment to a two-dimensional map, so that elevation information can be fully utilized.

2.7. The dynamic change of the weighting function can be considered to reflect real-time environment or traffic conditions, such as increasing the weighting of mountain roads when bad weather is encountered. In order to make the path planning more intelligent, safer and more practical, various environmental factors must be comprehensively considered, and the factors are reflected into the graph algorithm through a reasonably designed weight function.

The final function expression:

α, β, γ, δ are weight coefficients for balancing different factors.

TerrainWeight (u, v) represents terrain weights, which may be specified according to the type of terrain.

Geologicweight (u, v) represents a geological weight, which can be specified according to geological conditions (hard rock, soft soil, etc.).

Dangerweight (u, v) represents a risk weight, and the weight assigned to the side considered to be dangerous is manually specified to be 15%.

CostWeight (u, v) represents a running cost weight, and the time or energy cost under different environments can be learned by a machine learning algorithm.

3. Constraint conditions

Constraints are introduced, such as avoiding crossing dangerous areas or specific safety requirements. Aiming at the underground complex environment, on the basis of the traditional Dijkstra shortest path algorithm, the following constraint conditions are added:

3.1. and avoiding dangerous area constraint, namely identifying dangerous areas existing underground in advance, and enabling the path to bypass the dangerous areas in the path searching process.

3.2. Flatness constraint: preferably, a path with a smoother ground is selected to ensure that the unmanned vehicle runs stably.

3.3. Turning radius constraint: the turning radius of the path cannot be smaller than the minimum turning radius of the drone.

3.4. Vehicle body size constraints: the width of the path must be greater than the size of the body of the unmanned vehicle, avoiding jamming.

3.5. Visibility constraints: preferably, a path with good illumination condition and wide visual field is selected.

3.6. Ground constraint: the car body is always kept on the ground, so that suspension is avoided.

3.7. Carrying weight constraints: the path must be able to withstand the load requirements of the drone.

3.8. Mineral wall and pillar restraint: maintaining a safe distance from the mine wall and the support post.

By adding the constraint conditions, the method can plan a safer and smoother underground unmanned mine card navigation path. For dangerous area constraint, whether the edge passes through the dangerous area can be checked when the weight is calculated, and if so, the weight is reset to infinity; the flatness constraint can be realized by adding a term considering the flatness of the ground into the weight function; the turn radius constraint may be achieved by checking at the curvature of each segment on the path; the body size constraint may be implemented by taking the path width into account in the weight function; the visibility constraint may be implemented by adding a term that considers visibility in the weighting function; the ground constraint may be achieved by checking Gao Chenglai for each point on the path in the elevation map; the carrying weight constraint may be achieved by taking into account the carrying capacity of the path in the weight function; the mine wall and pillar constraints may be achieved by adding distance-considered terms to the weighting function.

Improved path planning algorithm framework:

initializing:

initializing a starting point and adding the starting point to a priority queue, wherein the initial path weight is 0.

Initializing the path weight of each node to infinity, indicating that it has not been accessed.

An empty path set is initialized.

Main cycle:

and taking out the node with the minimum current path weight from the priority queue.

For each neighbor of a node:

the original weights (distance or time, etc.) are calculated.

It is checked whether all the additional constraints are satisfied, and if so, the weights of the additional constraints are calculated.

The total weight, i.e. the original weight plus the weight of the additional constraint, is calculated.

If the new path weight is less than the currently known path weight, the path weight is updated and the path is updated.

If the node is not in the path set, it is added to the priority queue.

Termination condition:

the algorithm is stopped when the target node is added to the path set or the priority queue is empty.

And (3) path backtracking:

and backtracking from the end point to the start point to obtain the optimal path.

Algorithm modification points:

in calculating the original weights, the weights (distance, time, etc.) of the edges in the graph are used.

For each additional constraint, the weight function is modified according to the problem requirements and a corresponding check is made in the algorithm.

The path weights of the nodes are used in the priority queue for ordering.

In the case of dynamic weight adjustment, the weights are updated periodically or according to real-time data.

4. Obstacle avoidance and safety

Obstacle and hazard areas are detected using sensing technologies such as lidar, cameras and ultrasonic sensors. Meanwhile, the implementation of the obstacle avoidance algorithm is described to ensure the path safety. The sensing technology is utilized to realize path planning and obstacle avoidance of the underground unmanned mine card:

4.1. and carrying out three-dimensional scanning on the surrounding environment by using a laser radar to obtain point cloud data. Various barriers on the road surface, such as stones, pits and the like, can be detected through the point cloud data. And at the same time, the size and position information of the obstacle can be measured.

4.2. And acquiring pavement image information by using a camera. By the image processing feature extraction method, dangerous areas such as road surface damage, inclination, water accumulation and the like can be detected. The dangerous areas can be semantically segmented and classified in combination with Convolutional Neural Networks (CNNs).

4.3. The distance between the road surface edge and the cliff edge is measured using an ultrasonic sensor to determine the boundary of the safety area. The ultrasonic sensor has high ranging precision and is not influenced by environmental illumination and dust.

4.4. And fusing the data of various sensors to establish a road surface three-dimensional map and an obstacle information database. And marking out dangerous areas, and determining areas needing to be bypassed when planning the global path.

4.5. In path planning, taking security as priority, searching the global shortest path by utilizing the improved Dijkstra algorithm. The algorithm may set weights for each road segment, avoiding pre-marked hazard areas.

4.6. In the movement process, the laser radar and the camera are utilized to detect the obstacle in real time, the distance of the obstacle is determined by combining the ultrasonic waves, and the local obstacle avoidance algorithm is operated to adjust the path so as to avoid the passage of the obstacle to the greatest extent.

4.7. If an impending area is encountered, the unmanned mining truck may automatically change direction of movement, planning a new path along the edge of the obstacle to leave the obstacle area.

4.8. After the destination is reached, the unmanned mining truck can store the road surface three-dimensional map and the obstacle information constructed in real time for use in next navigation. The sensor is used for collecting data, planning a global path and adjusting a local path in real time, so that the underground unmanned mining card can safely and reliably reach a preset destination.

The mining card kinematic model is very important for unmanned mining card, however, the model expression modes derived in related documents are different, and even some of the models have errors. In order to build the follow-up work on a reliable basis, the present embodiment re-combs and derives the ore card kinematics model, with emphasis on the relationship between articulation angle, articulation angular velocity, vehicle body linear velocity, and vehicle body angular velocity.

FIG. 1 is a model turn description, with point P in the two front wheel links ₁ Point P in the connection line of two rear wheels ₂ The hinge point is denoted O. The distance from the front axle to the hinge point is denoted as L ₁ The distance from the rear axle to the hinge point is denoted as L ₂ . The articulation angle is positive in the counterclockwise direction, denoted as beta. The front half and the rear half of the vehicle are set up according to the above figures,its coordinate origin is P ₁ And P ₂ The horizontal and vertical lines represent the X-axis and Y-axis, respectively, and these two coordinate frames are denoted as sigma ₁ Sum sigma ₂ . Heading angle of the front half of the vehicle is recorded as theta ₁ The heading angle of the rear half of the vehicle is recorded as theta ₂ The heading angle is positive in the counterclockwise direction. The map coordinate system (which may be assumed to be the inertial system) in which the vehicle is located is Σ.

It is assumed that the front and rear wheels meet the usual non-integrity constraints, that is, that the tire does not slip. Under this assumption P ₁ And P ₂ The speeds of two points, which are respectively denoted as v, can only be in the horizontal X-axis direction passing through the point in the above graph ₁ And v ₂ . The forward speed direction is the same as the horizontal X-axis direction, and the reverse speed direction is opposite to the horizontal X-axis direction.

The following equation is not difficult to obtain according to the geometrical relationship between the front and rear vehicle bodies

θ ₁ ＝θ ₂ +β (1)

The derivatives of the two sides are obtained

The speed of the hinge point O in the map coordinate system is considered from the front and rear bodies, respectively. In accordance with the basic principle of rigid motion, the front vehicle body coordinate system sigma ₁ In which the velocity of the O-point is equal to P ₁ Speed of point plus O point around P ₁ The speed at which the point rotates. Projecting the above speeds to a map coordinate system may result in:

similarly, in the rear body coordinate system Σ ₂ In which the velocity of the O-point is equal to P ₂ Speed of point plus O point around P ₂ The speed at which the point rotates. The speed is projected to a map coordinate system to obtain

Since the formula (2) and the formula (3) are both the speeds of the O point in the map coordinate system, the two formulas are equal, and thus the result is

Both sides of the formula (4) are two-dimensional column vectors, so that the left multiplication vector (-sin theta) can be simultaneously used ₁ ,cosθ ₁ ) The product and difference formula of the trigonometric function is adopted to be properly arranged to obtain

Bringing formula (1 a) into formula (5) and properly arranging to obtain

Similarly, the left multiplication vector (-sin theta) is simultaneously formed on two sides of the formula (4) ₂ ,cosθ ₂ ) The integration and difference formulas of the trigonometric functions are adopted to be properly arranged and are carried into the formula (1 a), and the obtained product can be obtained after the arrangement

If the coordinate frame of the car body is sigma ₂ In the center, the kinematic model should be represented by formula (6). For convenience of comparison, use θ _r Representing the heading angle of the rear vehicle body and v representing the rear vehicle body P ₂ The linear velocity of the point, denoted by gamma, the articulation angle, denoted by L _r Representing the distance from the rear axle to the hinge point, using L _f Representing the distance of the front axle to the hinge point, then equation (6) can be written as

This section will describe the setup of the experiment, data collection and analysis of the results. The performance of the improved algorithm in path planning is demonstrated through simulation or testing in an actual downhole environment.

Experiment setting: experimental scenarios for testing the improved algorithm are described, including map data, configuration of unmanned mining cards, and environmental parameters.

The dimensions of the mining truck model of this embodiment are shown in fig. 2 as follows:

maximum total length: 1130, a. Tire diameter: 16.9, b. rear axle center to hinge chain center distance: 18.75, c. rear axle center to IMU center distance: distance between imu and right rear wheel center: 4.4, e. rear left right wheel center distance: 24, f. Front radar reaches front axis center distance: 6.8, g.wheelbase: 405, h. rear axis center to rear radar distance: 4.75, maximum turning radius: 53 deg..

Unmanned mining card navigation path planning experimental scene of improved Dijkstra algorithm is tested:

the experimental scene is set in an office with multiple ground positions, and the navigation map data comprises key position information such as a plurality of branches and channels, an entrance and an exit of an ore card model car and the like. The parameter configuration of the unmanned mining card comprises maximum speed, turning radius, carrying capacity and the like, and the unmanned mining card needs to move in complex terrains and complete the excavation task.

The environmental parameters are mainly:

terrain complexity: how many branches, curves and other factors affect the planning difficulty in the map;

terrain variation: whether the mine channel is considered to possibly deform to influence a feasible path or not;

traffic flow: the traffic flow of other load-carrying vehicles influences the path selection and the time; visibility: the visibility difference of different areas affects the path planning algorithm;

by setting map scenes with different difficulties, configuring different unmanned mining card body parameters and comprehensively considering the change of environmental factors, the difference between the Dijkstra algorithm and the original algorithm after the improvement of the test can be compared from multiple dimensions, and the planning efficiency and the path quality of the improved algorithm under the complex scene can be verified. This allows a comprehensive assessment of the performance improvement of the improved algorithm.

And (3) data collection: the data collection process is described in detail, including sensor data, path planning results, and data related to improving algorithm performance.

Sensor data collection:

performing regional scanning by using a laser radar to obtain well wall surface point cloud data;

acquiring position and posture data of the unmanned vehicle by using an inertial navigation system;

acquiring cargo box load information by using a load sensor;

and acquiring container position information by using a mechanical arm position sensor.

Path planning:

processing the point cloud data into a grid map, and marking the positions of the obstacles;

marking a starting point, an ending point and a charging station position on a map;

based on the improved Dijkstra algorithm, carrying out shortest path planning on a map;

the path planning result comprises a global path route and local obstacle avoidance information.

Improved algorithm performance evaluation data:

recording path planning time;

comparing the path length of the improved algorithm with that of the original algorithm;

counting the obstacle avoidance times;

recording the repeated planning times;

calculating the smoothness of the path;

evaluating the safety and feasibility of the planned path;

through the collection of the data, the effect of improving the Dijkstra algorithm in a downhole environment can be evaluated, wherein the improvement comprises the improvement of algorithm speed, path quality and the like. These data are helpful for further optimization algorithms, improving autonomous navigation capabilities of the drone.

Analysis of results: and analyzing experimental results, wherein the experimental results comprise results in the aspects of performance comparison of an improved algorithm and a traditional Dijkstra algorithm, path planning efficiency, safety, energy consumption optimization and the like. And performing path searching. Compared with the traditional Dijkstra algorithm, the improved algorithm has the following advantages:

the improved algorithm considers the constraint of turning radius and can generate a travelable path meeting the characteristics of the unmanned mining card body. The path generated by the conventional Dijkstra algorithm may have a tight turn that is not feasible.

The improved algorithm adds a comprehensive evaluation function in the edge weight calculation, so that the generated path is shortest in distance, and the safety and smoothness of the path are considered. The traditional Dijkstra only targets the shortest distance.

The improved algorithm adopts a better priority queue structure, and reduces the time complexity of single extraction of the minimum path overhead. This results in an improvement in the overall time efficiency of the algorithm.

The improved algorithm uses the destination guidance strategy in the searching process, so that the relative optimal path in the destination direction can be searched out more quickly, and the total searching quantity is reduced.

Table 1 comparison of results before and after improvement

	Average generation time(s)	Average number of turns	Average path length	Counting times
					Traditional algorithm	2.2	6.4	20.6	192
Improved algorithm	1.83	5.8	20	205
					Improvement of ratio of merit	16.7	9.4	3.2

As shown in Table 1, comparison of simulation experiments shows that under the same scene condition, the average length of the paths generated by improving Dijkstra algorithm is 3.2% shorter than that of the paths generated by the traditional algorithm, the calculation time is reduced by 16.7%, the turning times are reduced by 9.4%, and the ratio of the paths meeting the motion constraint of the unmanned mining card is improved by 5.2%.

This shows that the improved algorithm can achieve a better path planning effect.

In general, the improved Dijkstra algorithm remarkably improves the navigation path planning efficiency of the underground unmanned mining card, so that the generated path meets the actual requirements, unsafe driving behaviors are reduced, the service life of equipment is prolonged, the operation efficiency is improved, and the method is worthy of popularization and application.

In order to achieve the above method, the embodiment of the application also provides an unmanned driving path planning system for the underground mine truck, which comprises a laser radar, an inertial navigation system, a load sensor, a mechanical arm position sensor and a processor connected with the sensor, wherein the processor is also connected with a display and a memory, and the laser radar performs regional scanning to obtain the cloud data of the wall surface points of the well; the inertial navigation system acquires position and posture data of the unmanned vehicle; the load sensor acquires cargo information of a cargo box; the mechanical arm position sensor obtains container position information. The processor performs unmanned driving path planning of the mine truck based on the acquired data, and the laser radar, the inertial navigation system, the load sensor, the mechanical arm position sensor, the processor, the display and the memory hardware are all existing products.

The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.

Claims

1. An improved Dijkstra underground mine truck unmanned path planning method is characterized in that: comprising the following steps:

step (1) data collection

step (2) Path planning

2. The method according to claim 1, characterized in that: processing the point cloud data into node and side information, using key points in the feature extraction environment as nodes, and calculating the distance and direction between the nodes as the side information by using the node data; as unmanned mining trucks move in the environment, the newly acquired sensor data is continually used to optimize and enhance the accuracy of the map; adopting a graph optimization algorithm to continuously adjust the node positions and delete the error nodes and edges so that the map is closest to the real environment; when the unmanned mining truck returns to the previously traversed place, loop detection is realized through feature matching, so that the accuracy of the map is further improved, and the node positioning is optimized.

3. The method according to claim 1, characterized in that: in the modified Dijkstra algorithm, the weight function is as follows:

α, β, γ, δ are weight coefficients for balancing different factors;

geologicweight (u, v) represents geological weights, and is specified according to geological conditions;

dangerweight (u, v) represents a risk weight, manually assigning a weight of 10-20% to the edge considered to be dangerous;

4. The method according to claim 1, characterized in that: in the modified Dijkstra algorithm, the constraint conditions are as follows:

5. The method according to claim 1, characterized in that: the improved path planning algorithm framework includes:

6. An unmanned route planning system of a mine card in pit, which is characterized in that: the system comprises a laser radar, an inertial navigation system, a load sensor, a mechanical arm position sensor and a processor connected with the sensors, wherein the processor is also connected with a display and a memory, and the laser radar performs regional scanning to acquire wall surface point cloud data of a well; the inertial navigation system acquires position and posture data of the unmanned vehicle; the load sensor acquires cargo information of a cargo box; the mechanical arm position sensor acquires container position information; the processor performs mine truck unmanned route planning based on the acquired data, and the method is performed according to any one of claims 1 to 5.

7. An electronic device, characterized in that: computer program comprising a memory, a processor and a computer program on the memory and executable on the processor, said processor implementing the steps of the method according to any of the preceding claims 1 to 5 when said computer program is executed.

8. A non-transitory computer readable storage medium characterized by: a computer program stored thereon, which, when executed by a processor, implements the steps of the method according to any of claims 1 to 5.