CN110513120A - Self-adaptive positioning system and method for cutting head of roadheader - Google Patents

Abstract

The invention discloses an adaptive positioning system and method for a cutting head of a roadheader based on optical and magnetic field positioning. The magnetic field positioning part in the present invention first traverses the working space through the cutting head with a fixed permanent magnet, the optical positioning module records the three-dimensional space position in real time, and the magnetic field sensor records the magnetic induction intensity of the corresponding position to obtain training data, and then passes The deep learning algorithm constructs the initial magnetic field positioning model. During the working process of the magnetic field positioning model, the optical auxiliary positioning system continues to work, and keeps collecting the position information of the cutting head when the cutting head of the roadheader is not buried. The positioning model is updated for training. During the update training process, the data provided by the optical-assisted positioning results has a time weight. The closer the data is to the current time, the greater the learning rate in the model update training. Through the continuous updating and training of the magnetic field positioning model, it can automatically adapt to environmental changes, avoiding the impact of demagnetization of the magnetic source due to vibration and geomagnetic changes after long-distance movement of the roadheader on the magnetic field positioning accuracy.

Description

Self-adaptive positioning system and method for cutting head of roadheader

technical field

The invention relates to a positioning system and method for a cutting head of a roadheader. Specifically, the present invention relates to a cutting head positioning system and method of a roadheader that can automatically adapt to environmental changes based on deep learning-based magnetic field positioning, with optical positioning as an auxiliary calibration.

Background technique

It is a feasible method to use the magnetic field positioning method to locate the cutting head of the roadheader in the environment of high dust and low visibility. However, the magnetic field positioning method also has its limitations.

Aiming at the problem that the accuracy of the magnetic field positioning system of the cutting head of the roadheader decreases with time, that is, the long-term and large-scale movement of the roadheader causes changes in geomagnetic conditions, and the magnetic source located at the back of the cutting head of the roadheader due to long-term Due to the decrease in magnetic field positioning accuracy caused by the weakening of vibration and magnetism, we introduced optical positioning for auxiliary calibration, and made dynamic improvements to the magnetic field positioning model, realizing the automatic calibration adjustment of the cutting head positioning system of the roadheader, so that it can automatically Adapt to environmental changes, maintain positioning accuracy in a dynamic environment, and greatly improve the stability of the cutting head positioning system of the roadheader.

Contents of the invention

The invention discloses an adaptive positioning system and method for the cutting head of a roadheader, which solves the problem that the positioning accuracy decreases as the positioning system changes with the environment and the magnetic source intensity attenuates during the positioning process of the roadheader by using the magnetic field positioning method , greatly improving the stability of the cutting head positioning system of the roadheader. The adaptive positioning system includes a magnetic field positioning module, a magnetic source that can be fixed behind the cutting head, an optical auxiliary positioning module, and a terminal processor containing an adaptive positioning model. Due to the large-scale movement of the roadheader and the passage of time, the geomagnetic conditions will change accordingly. Due to the impact of the vibration of the cutting head, the strength of the magnetic source at the rear of the cutting head will also be affected, based on the initial geomagnetic environment and The accuracy of the magnetic field positioning model on the strength of the magnetic source will inevitably decrease or even fail. The magnetic field positioning part in this method first traverses the working space through the cutting head with a fixed permanent magnet, the three-dimensional space position is recorded in real time by the optical auxiliary positioning module, and the magnetic induction intensity of the corresponding position is recorded by the magnetic field sensor to obtain training data, and then Build a magnetic field positioning model through machine learning algorithms. Then, during the working process of the magnetic field positioning model, the optical assisted positioning system continues to work, and keeps collecting the position information of the cutting head when the cutting head of the roadheader is not buried, and the high-precision position information collected by the optical assisted positioning system The magnetic field positioning model is updated and trained. During the update training process, the data provided by the optical aided positioning results has a time weight. The closer the data is to the current time, the greater the learning rate in the model update training. Through the continuous updating and training of the magnetic field positioning model, it can automatically adapt to environmental changes, avoiding the impact of demagnetization of the magnetic source due to vibration and geomagnetic changes after long-distance movement of the roadheader on the magnetic field positioning accuracy.

The specific technical contents provided by the invention are:

An adaptive positioning system for a cutting head of a roadheader. The positioning system includes a magnetic field positioning module, a magnetic source that can be fixed behind the cutting head, an optical auxiliary positioning module, and a terminal processor including an adaptive positioning model. The magnetic field positioning module is used to perform magnetic field positioning on the cutting head; the magnetic source is used to transmit the position data of the cutting head to the magnetic field positioning module; the optical auxiliary positioning module is used to provide high precision required for calibration and adjustment data; the terminal processor containing the adaptive positioning model is used to process the obtained position data, and automatically adjust the magnetic field positioning model according to environmental changes to maintain its positioning accuracy; the position data includes high-precision position data obtained by the auxiliary positioning module , and the magnetic field position data obtained by the magnetic field positioning module.

Preferably, the magnetic field positioning module includes at least two three-axis magnetic field sensors, which are respectively placed on both sides of the body of the roadheader or on slide rails and supports that move synchronously with the body.

Preferably, the optical auxiliary positioning module includes at least two optical cameras and supporting data acquisition and transmission devices. There will be no change in the above, the optical camera is not limited to the visible light band, and cameras of various bands such as far and near infrared can be used according to the specific environment.

Preferably, the magnetic source is fixed behind the cutting head and moves together with the cutting head, and the magnetic field intensity obtained by the magnetic field sensor is affected by the movement of the magnetic source. The magnetic source is a permanent magnet or an electromagnet.

The positioning method disclosed in the present invention is as follows: the magnetic field positioning method utilizes the machine learning algorithm in the processor to train the samples to build a positioning model, and uses the positioning model to perform magnetic field positioning on the cutting head; the machine learning algorithm is depth Learning algorithm, the positioning model is a deep learning positioning model, and the deep learning positioning model is obtained after training with sample data. The deep learning positioning model is used for magnetic field positioning of the cutting head, and the magnetic field corresponding to different positions is obtained by a magnetic field sensor. The magnetic field strength data of the source, the deep learning positioning model uses the magnetic field strength data to obtain the spatial position of the cutting head. The optical-assisted positioning method locates the cutting head based on image recognition and geometric optics. The image recognition method can be based on the imaging characteristics of the cutting head under visible light conditions, or can be generated based on the frictional heating of the cutting head in operation. Infrared imaging features; the automatic calibration method updates and trains the model used for magnetic field positioning according to the results of optical assisted positioning. The larger the learning rate of the data in the model update. During the positioning process of the cutting head, the magnetic field positioning is used to provide the positioning information of the cutting head globally and at all times. The optical auxiliary positioning system only works when the cutting head of the roadheader is not blocked by the fuselage or buried in the slag, providing high The precision position data is used to automatically adjust and calibrate the magnetic field positioning model to avoid the influence of magnetic source demagnetization due to vibration and geomagnetic changes after long-distance movement of the roadheader on the magnetic field positioning accuracy.

Preferably, the automatic calibration method includes the following steps:

Step 1, the optical auxiliary positioning system obtains the position information of the cutting head;

Step 2, the magnetic field sensor collects magnetic field data;

Step 3, judging whether the optical positioning system is in a normal working state. If yes, go to step 4; if not, go to step 10;

Step 4, store the optical positioning result into the historical data, data serial number i;

Step 5, store the magnetic field data into the historical data, the data sequence number is i;

Step six, i=i+1;

Step 7, update the historical data time weight, the greater the serial number i, the greater the weight;

Step 8, adjust the learning rate of different time data when updating the training model according to the time weight;

Step 9, using historical data to update the training magnetic field positioning model;

Step ten, bring the magnetic field data acquired by the magnetic field sensor into the magnetic field positioning model;

Step eleven, outputting the magnetic field positioning result.

Description of drawings

Fig. 1 schematic diagram of simulation model;

Fig. 2 schematic diagram of magnetic field signal and corresponding positioning coordinates;

Fig. 3 introduces a schematic diagram of a magnetic field change;

Fig. 4 Flowchart of adaptive positioning algorithm;

Fig. 5 Ordinary magnetic field positioning model positioning effect diagram of original data;

Figure 6. The effect diagram of the positioning effect of the common magnetic field positioning model on the data introduced by the magnetic field change;

Fig. 7 The localization effect diagram of the adaptive model for the data introduced with the change of the magnetic field.

Detailed ways

The present invention will be described in further detail through implementation examples below in conjunction with the accompanying drawings, but the scope of the present invention is not limited in any way.

In the implementation cases of the present invention, the simulation data is used to demonstrate the positioning effect of the self-adaptive positioning model.

As shown in Figure 1, in the simulation model, the magnetic source is a 500-turn circular electromagnet with an inner diameter of 55cm and an outer diameter of 60cm, and the working current is 1A. Located 500cm above the plane where the sensor is located. The magnetic field sensor is located in the center of the plane in which the sensor is located.

The invention provides a composite positioning system for cutting head of roadheader based on infrared and magnetic field positioning. The positioning system includes an infrared positioning module, a magnetic field positioning module, and a terminal processor containing an adaptive positioning model. The simulation data used in the specific embodiment is specifically shown The localization effect of the adaptive localization model.

The above is the specific composition of the cutting head positioning system of the roadheader of the present invention. The positioning method of the present invention will be described in detail below.

Magnetic field positioning requires pre-training positioning model. In this embodiment, firstly, the cutting head with fixed permanent magnet traverses the working space, the infrared positioning module records the three-dimensional space position from time to time, and the magnetic field sensor records the magnetic induction intensity of the corresponding position to obtain training data. In the simulation model, we can directly obtain and use the actual position of the magnetic source and the magnetic field collected by the magnetic field sensor. The data obtained by the simulation is shown in Figure 2. The coordinate information obtained by the simulation is used to replace the positioning information of the optical assisted positioning, and the magnetic field information obtained by the simulation is used to replace the magnetic field information collected by the magnetic field sensor in the actual positioning. We simulate the change of geomagnetic field by introducing a magnetic field change that changes with the positioning sequence to show the positioning effect of the adaptive positioning model and the common model under the condition of geomagnetic change. The magnetic field change introduced is shown in Figure 3. In the adaptive localization model, in addition to the initial training, we train the localization model in real time to ensure the localization accuracy after the magnetic field changes are introduced.

The flow chart of the above adaptive positioning algorithm is shown in FIG. 4 . The optimization algorithm includes the following steps:

Step 1, the optical auxiliary positioning system obtains the position information of the cutting head;

Step 2, the magnetic field sensor collects magnetic field data;

Step 3, judging whether the optical positioning system is in a normal working state. If yes, go to step 4; if not, go to step 10;

Step 4, store the optical positioning result into the historical data, data serial number i;

Step 5, store the magnetic field data into the historical data, the data sequence number is i;

Step six, i=i+1;

Step 7, update the historical data time weight, the greater the serial number i, the greater the weight;

Step 8, adjust the learning rate of different time data when updating the training model according to the time weight;

Step 9, using historical data to update the training magnetic field positioning model;

Step ten, bring the magnetic field data acquired by the magnetic field sensor into the magnetic field positioning model;

Step eleven, outputting the magnetic field positioning result.

The relative positioning error of the original positioning model without using the adaptive positioning algorithm is shown in Figure 5. It can be seen from Figure 5 that the relative error is less than 1.6%. Ordinary positioning models are highly accurate.

The relative positioning error of the ordinary positioning model that does not use the adaptive positioning algorithm to locate the data after the magnetic field change is introduced is shown in Figure 6. It can be seen from Figure 6 that after the magnetic field change is introduced, that is, when the geomagnetism is dynamic, with time Over time, the relative positioning error of the common positioning model gradually increases. After the serial number is greater than 2000, the relative positioning error of the x-axis has exceeded 100%, and the common positioning model becomes invalid.

Using the positioning model of the adaptive positioning algorithm to locate the data after the magnetic field change is introduced, the positioning relative error is shown in Figure 7. It can be seen from Figure 7 that after the magnetic field change is introduced, that is, when the geomagnetic The relative positioning error of the adaptive positioning model has always been kept below 2.5%, and there is no phenomenon that the relative positioning error increases with time, which proves that the adaptive positioning model can well adapt to the gradually changing geomagnetic environment and maintain a high positioning accuracy.

It should be noted that the purpose of the disclosed embodiments is to help further understand the present invention, but those skilled in the art can understand that various replacements and modifications are possible without departing from the spirit and scope of the present invention and the appended claims of. Therefore, the present invention should not be limited to the content disclosed in the embodiments, and the protection scope of the present invention is subject to the scope defined in the claims.

Claims (8)

1. a kind of cutting head of roadheader adaptive location system and method, it is characterised in that:

The adaptive location system includes Magnetic oriented module, the magnetic source for being securable to cutterhead rear, optics auxiliary positioning Module, the terminal handler containing adaptive location model；The Magnetic oriented module is used to carry out Magnetic oriented to cutterhead； The magnetic source is used to the position data of cutterhead passing to Magnetic oriented module；The optics auxiliary positioning module is for providing High accuracy data needed for calibration adjustment；The terminal handler containing adaptive location model is used for acquisition position data It is handled, and according to environmental change adjust automatically Magnetic oriented model to keep its positioning accuracy；The position data packet The magnetic field position data that the high precision position data and Magnetic oriented module for including auxiliary positioning module acquisition obtain.

2. positioning system according to claim 1, which is characterized in that the Magnetic oriented module includes at least two 3 axis Magnetic field sensor is respectively placed on machine body of boring machine two sides or the sliding rail and bracket that move with body synch.

3. positioning system according to claim 1, which is characterized in that the optics auxiliary positioning module includes at least two Optical camera and the acquisition of matched data, transmitting device, at least two optical camera are located at cutterhead rear, and After being fixed with fuselage relative positional relationship, do not rechanged in position fixing process, the optical camera is not limited to visible light wave Section can use the camera of far and near a variety of wave bands such as infrared according to specific environment.

4. positioning system according to claim 3, which is characterized in that magnetic source is fixed on cutterhead rear, with cutterhead one Movement is played, the magnetic field strength that the magnetic field sensor obtains is influenced by the movement of magnetic source；The magnetic source is permanent magnet or electromagnetism Iron.

5. a kind of cutting head of roadheader adaptive location method, using the described in any item positioning systems of claim 1-4 to pick Positioned into machine cutterhead, it is characterised in that: the adaptive location method include Magnetic oriented and optics auxiliary positioning and Automatic calibration；The Magnetic oriented method constructs positioning mould after being trained using the machine learning algorithm in processor to sample Type carries out Magnetic oriented to cutterhead using the location model；The machine learning algorithm is deep learning algorithm, described fixed Bit model is deep learning location model, and deep learning location model, the depth are obtained after being trained using sample data It practises location model to be used to carry out Magnetic oriented to cutterhead, the magnetic of the magnetic source at corresponding different location is obtained by magnetic field sensor Field intensity data, the deep learning location model obtain the spatial position of cutterhead using the magnetic field strength date；It is described Optics assisted location method is based on image recognition and method of geometrical optics positions cutterhead, and described image recognition methods can With the imaging features based on cutterhead under visible light conditions, can also be generated based on the frictional heating of the cutterhead in work red Outer imaging features；The automatic calibrating method is updated model used in Magnetic oriented according to optics auxiliary positioning result It trains, in the update training process, data provided by optics auxiliary positioning result have time weighting, apart from current time Closer data learning rate in model modification is bigger；Magnetic oriented is mentioned for global all the period of time in the position fixing process of cutterhead For cutterhead location information, optics aided positioning system is not only blocked by fuselage in cutting head of roadheader or is not buried in slag Work when middle, provide high precision position data to adjust automatically calibrate Magnetic oriented model, avoid magnetic source because concussion demagnetization with And development machine move at a distance after influence of the GEOMAGNETIC CHANGE to Magnetic oriented precision.

6. adaptive location method according to claim 5, it is characterised in that: the Magnetic oriented algorithm is deep learning Algorithm obtains the high deep learning location model of positioning accuracy after being trained using the optimization algorithm of deep learning.

7. adaptive location method according to claim 5, it is characterised in that: the optics assisted location method is by image The three-dimensional location of recognizer and method of geometrical optics is constituted.

8. adaptive location method according to claim 5, it is characterised in that: the automatic calibration process is auxiliary according to optics It helps positioning to provide high precision position information and adjusts the Magnetic oriented model obtained by deep learning, the adjustment process is to magnetic Field location model is trained again, the training process again real-time perfoming, institute under optics auxiliary positioning normal running conditions It states in training process again, the high precision position information that the optics auxiliary positioning closer apart from current time provides has higher Weight, the weight is to determine the learning rate size in training process again, and weight is higher, and learning rate is bigger.