Intelligent coal mine vehicle dispatching control method based on position feedback

Through multimodal sensors and improved reinforcement learning algorithms, high-precision position feedback and dynamic scheduling of coal mine vehicles are achieved, solving the problems of inaccurate positioning and low resource utilization efficiency in traditional scheduling, improving scheduling efficiency and security, and ensuring Continuity and safety of production.

CN119045443AInactive Publication Date: 2024-11-29陕西智引科技有限公司

Patent Information

Application Number
CN202411523618.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-30
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Traditional coal mine vehicle dispatching relies on manual experience and simple communication tools, making it difficult to accurately grasp the vehicle position, resulting in inaccurate driving path planning, low resource utilization efficiency, lack of effective fault monitoring and emergency response mechanisms, and being unable to respond to emergencies quickly, affecting production efficiency and safe.

Method used

Multimodal sensors are used to collect vehicle position information, use the improved Kalman filtering algorithm to achieve high-precision position feedback control, dynamically adjust driving paths and task allocation based on the improved reinforcement learning algorithm, issue scheduling instructions in real time through wireless communication and monitor execution, and automatically execute Emergency control, including fault isolation and path re-planning.

Benefits of technology

It realizes high-precision positioning and dynamic scheduling, reduces vehicle air driving and congestion, improves scheduling efficiency and resource utilization, ensures production continuity and safety, responds quickly to emergencies, automatically responds to failures, and improves overall production efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure SMS_75
    Figure SMS_75
  • Figure SMS_76
    Figure SMS_76
  • Figure SMS_77
    Figure SMS_77
Patent Text Reader

Abstract

The invention belongs to the technical field of coal mine vehicle dispatching control, and particularly relates to an intelligent dispatching control method based on position feedback. Firstly, position information is collected through multiple positioning technologies, and high-precision position feedback control is achieved through an improved Kalman filtering algorithm. And dynamically adjusting a vehicle driving path and a task allocation strategy based on an improved reinforcement learning algorithm. And then a scheduling instruction is issued through a wireless communication network, and monitoring optimization is carried out. And finally, vehicle faults and emergencies are detected, and emergency control including fault isolation and path re-planning is automatically executed. According to the method, the positioning precision is improved through the multi-modal sensor, and the limitation of a single sensor is overcome. An improved reinforcement learning algorithm improves scheduling efficiency, emergency control guarantees production continuity and safety, and many problems of a traditional coal mine vehicle scheduling method are effectively solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of coal mine vehicle dispatching control, and in particular relates to an intelligent dispatching control method based on position feedback. Background Art

[0002] In coal mine production operations, vehicle dispatching is a key link to ensure efficient and safe production. Traditional coal mine vehicle dispatching mainly relies on manual experience and simple communication tools. There are many limitations in manual dispatching. First, it is difficult to accurately grasp the real-time location of the vehicle, which makes the vehicle driving path planning lack of accuracy, and it is easy to have empty vehicles, congestion, etc., which seriously affects the transportation efficiency. Secondly, the task allocation is not reasonable enough, and it cannot be dynamically adjusted according to the actual status and working environment of the vehicle, resulting in some vehicles being overloaded with tasks, while some vehicles are idle, and resource utilization efficiency is low. Furthermore, for vehicle failures and emergencies, the existing dispatching method lacks an effective monitoring and emergency response mechanism. When a vehicle fails, the failure cannot be discovered and isolated in time, which may cause a safety accident. Moreover, in the face of sudden changes in road conditions or other emergencies, it is impossible to quickly replan the vehicle path, further affecting the continuity of production. With the development of the coal mining industry, the requirements for production efficiency and safety are increasing day by day. The traditional dispatching method can no longer meet the needs. There is an urgent need for an intelligent and precise vehicle dispatching control method to solve these problems. Summary of the invention

[0003] In view of the above-mentioned technical problems, the present invention proposes an intelligent scheduling control method based on position feedback.

[0004] In order to achieve the above object, the technical solution adopted by the present invention comprises the following steps:

[0005] S1. First, the location information of coal mine vehicles is collected through a variety of positioning technologies, and the improved Kalman filter algorithm is used to achieve high-precision position feedback control;

[0006] S2. Secondly, based on the improved reinforcement learning algorithm, the vehicle driving path and task allocation strategy are adjusted dynamically in real time to ensure the efficient execution of the scheduling task;

[0007] S3, then issue dispatch instructions in real time through the wireless communication network, and monitor and optimize the execution of the vehicle through position feedback;

[0008] S4. Finally, vehicle failures and emergencies are detected through vehicle status monitoring and safety strategies, and emergency control is automatically performed, including fault isolation and path replanning;

[0009] The implementation of the emergency control in step S4 includes:

[0010] S41, firstly, use intelligent control logic to achieve fault isolation, determine whether there is a fault based on historical data and fault threshold settings, and enable safety mode and restrict vehicle movement when a fault occurs;

[0011] S42, Secondly, for path replanning combined with multi-path selection strategy, under normal circumstances, the dynamic programming algorithm is used to generate the initial path ,in is the cost of each node on the path, and m is the total number of nodes; multiple alternative paths are selected based on the preliminary path, and the real-time cost of each alternative path is calculated; finally, the optimal path is selected based on the real-time evaluation.

[0012] Preferably, the implementation steps of step S1 are:

[0013] S11, first use GPS, INS, UWB, RFID multi-modal sensors to collect the initial position information of the mine car, and initialize the state , the initial covariance matrix and the measurement noise covariance matrix for each sensor and the process noise covariance matrix ;

[0014] S12. Secondly, based on the real-time performance data of the sensor, an adaptive noise estimation method is used to estimate the measurement noise covariance matrix of each sensor. , where k is the time, for the observation data of sensor i , update the noise level by measuring the error: ,in is the smoothing coefficient, is the residual, where is the observation matrix, is the state estimate vector;

[0015] S13. Then, the data fusion of each sensor is dynamically weighted. The weight is proportional to the inverse of the noise covariance. The smaller the noise of the sensor, the greater its weight, and vice versa. The weight of each sensor is calculated as ,in To measure the trace of the noise covariance matrix, the weights are normalized to ensure that the sum of the weights of all sensors is 1;

[0016] S14, then use the weighted observation value of each sensor To correct the state prediction ,in is the Kalman gain;

[0017] S15. Finally, use the new weights and Kalman gain to update the state covariance matrix , where I is the identity matrix.

[0018] Preferably, the operation of using improved reinforcement learning to achieve real-time dynamic adjustment of the vehicle's driving path and task allocation strategy in step S2 is:

[0019] S21. First, define the state space S, and define each state as the vehicle's current location information, current task, and surrounding environment information. , define the action space ;

[0020] S22. Next, a reward function is designed to evaluate the benefits of each state-action pair, taking into account factors such as task completion rate, safety of the driving path, and minimum energy consumption. The reward function is ,in are weight coefficients, which represent the importance of different factors in the reward function;

[0021] S23, then use the improved method to update the Q value, the specific formula is: ,in is the learning rate, is the discount factor, is the updated Q value, is the Q value, indicating that in the state Take action The estimated long-term return of For the status The maximum Q value of all possible actions represents the maximum future reward that can be obtained in the next step;

[0022] S24, then adjust the weight coefficient in the reward function through the adaptive mechanism , dynamically optimize according to the actual situation of task execution ,in It is the result of feedback adjustment based on the difference between the current task execution efficiency and the expected value. is the adjustment rate;

[0023] S25. Finally, combined with model predictive control, the optimal state of the next step is predicted through the model of the current state to improve efficiency , where MPC is model predictive control.

[0024] Preferably, the method for issuing dispatch instructions in real time and monitoring and optimizing vehicle execution through position feedback in step S3 is as follows: first, a low-power wide area network architecture is established, and the dispatch center generates dispatch instructions according to the current status of the vehicle and the task priority; then the dispatch instructions are sent to each vehicle through the wireless network, and the message queue is used to ensure the timeliness and reliability of the instructions; then each vehicle regularly collects its own position information and sends it back to the dispatch center through the wireless network, and the vehicle position information is uploaded in real time to form a closed-loop monitoring; finally, the dispatch center receives the vehicle's position information and compares it with the predetermined path, evaluates the execution status and calculates the optimization index; if deviation delay is found, the dispatch center adjusts the dispatch instruction in real time and reissues it.

[0025] Compared with the prior art, the advantages and positive effects of the present invention are that it uses multimodal sensors to collect position information, overcomes the limitations of a single sensor through a series of processing, and achieves high-precision positioning in a complex coal mine environment. Based on an improved reinforcement learning algorithm, the state and action space are fully defined, a comprehensive reward function is designed, the Q value is updated, the weight coefficient is adaptively adjusted, and combined with model predictive control, the vehicle can quickly and accurately select the optimal path and task strategy, reduce empty driving and congestion, and improve overall scheduling efficiency. Automatic emergency control, fault isolation uses intelligent logic, judges faults based on data and thresholds, and enables safety mode to restrict vehicles; path replanning combines strategies, uses algorithms to generate paths and select the optimal path when normal, improves reliability and flexibility, and ensures production continuity and efficiency. DETAILED DESCRIPTION

[0026] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described below in conjunction with embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0027] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments of the following disclosure.

[0028] Example 1. In coal mine production operations, vehicle scheduling is crucial to the efficiency and safety of production. In order to improve the efficiency and safety of coal mine vehicle scheduling and solve the problems of task delays and resource waste caused by environmental complexity and sudden failures, an intelligent scheduling control method based on a multimodal sensor network is adopted. Through real-time monitoring, path replanning, task redistribution and other technical means, efficient management of coal mine vehicles is achieved.

[0029] In order to achieve comprehensive monitoring of coal mine vehicles, the present invention collects environmental data and vehicle status information in real time through a multimodal sensor network. The sensor network can provide high-precision data in a complex coal mine environment, ensuring that the actual position and status of the vehicle can be accurately determined in a variety of situations. First, GPS, INS, UWB, and RFID multimodal sensors are used to collect the initial position information of the mine vehicle and initialize the status. , the initial covariance matrix and the measurement noise covariance matrix for each sensor and the process noise covariance matrix ; Secondly, based on the real-time performance data of the sensor, the measurement noise covariance matrix of each sensor is estimated using an adaptive noise estimation method , where k is the time, for the observation data of sensor i , update the noise level by measuring the error: ,in is the smoothing coefficient, is the residual, where is the observation matrix, is the state estimation vector; then the data fusion of each sensor is dynamically weighted, and the weight is proportional to the inverse of the noise covariance. The smaller the noise of the sensor, the greater its weight, and vice versa. The weight of each sensor is calculated as ,in To measure the trace of the noise covariance matrix, the weights are normalized to ensure that the sum of the weights of all sensors is 1; then the weighted observations of each sensor are used To correct the state prediction ,in is the Kalman gain; finally, the state covariance matrix is ​​updated using the new weights and Kalman gain , where I is the unit matrix. This combination of multiple sensors overcomes the limitations of a single sensor and thus improves positioning accuracy.

[0030] Considering that the conventional reinforcement learning method adopts a more conventional Q-value update formula, it is not optimized for the characteristics of coal mine vehicle scheduling, and the weight coefficient is relatively fixed, and the adjustment method is relatively simple, which cannot adapt well to the actual changes in the vehicle scheduling process. The present invention adjusts the weight coefficient in the reward function through an adaptive mechanism. The weight coefficient is dynamically optimized according to the actual situation of task execution, and feedback adjustment is performed based on the difference between the current task execution efficiency and the expected value, and the adjustment rate is. This adaptive adjustment method can make the reward function more in line with the actual needs of the vehicle in different tasks and environments, thereby more effectively guiding the vehicle scheduling behavior. First, define the state space S, and define each state as the vehicle's current position information, current task, and surrounding environment information. , define the action space ; Secondly, a reward function is designed to evaluate the benefits of each state-action pair, taking into account factors such as task completion rate, safety of the driving path, and minimum energy consumption. The reward function is ,in are weight coefficients, which respectively indicate the importance of different factors in the reward function; then the improved method is used to update the Q value, the specific formula is: ,in is the learning rate, is the discount factor, is the updated Q value, is the Q value, indicating that in the state Take action The estimated long-term return of For the status The maximum Q value of all possible actions represents the maximum future reward that can be obtained in the next step;

[0031] Then the weight coefficient in the reward function is adjusted through the adaptive mechanism , dynamically optimize according to the actual situation of task execution ,in It is the result of feedback adjustment based on the difference between the current task execution efficiency and the expected value. is to adjust the rate; finally, combined with model predictive control, the optimal state of the next step is predicted through the model of the current state to improve efficiency , where MPC is model predictive control. The improvement of the present invention enables the vehicle to select the optimal driving path and task allocation strategy more quickly and accurately, reduce vehicle idling and congestion, and improve the overall dispatching efficiency of coal mine vehicles by comprehensively and accurately defining the state space and action space, as well as improving the reward function and Q value update method.

[0032] Next, in order to quickly adjust the driving path in an emergency situation and ensure that the vehicle can bypass obstacles or congestion areas, multiple paths are evaluated to improve the flexibility and safety of scheduling and reduce transportation delays. The present invention is based on a low-power wide area network architecture. The dispatch center generates a dispatch instruction based on the current state of the vehicle and the task priority; then the dispatch instruction is sent to each vehicle through a wireless network, and a message queue is used to ensure the timeliness and reliability of the instruction; then each vehicle regularly collects its own location information and sends it back to the dispatch center through a wireless network, and the vehicle location information is uploaded in real time to form a closed-loop monitoring; finally, the dispatch center receives the vehicle's location information and compares it with the predetermined path, evaluates the execution status and calculates the optimization index; when deviation delay is found, the dispatch center adjusts the dispatch instruction in real time and reissues it. This mechanism can respond quickly in an emergency situation. When encountering obstacles or congestion areas, multiple paths can be evaluated based on the real-time location information of the vehicle, and the optimal route can be selected to enable the vehicle to quickly bypass the obstacle, thereby improving the flexibility of scheduling. At the same time, through closed-loop monitoring and timely adjustment of instructions, the safety of vehicle driving is guaranteed, and safety accidents caused by path problems are avoided. It also reduces transportation delays and improves the overall efficiency of coal mine production.

[0033] Finally, when vehicle failures and emergencies are detected, emergency control is automatically performed, including fault isolation and path replanning. First, intelligent control logic is used to implement fault isolation, and whether a fault occurs is determined based on historical data and fault threshold settings. When a fault occurs, the safety mode is enabled and the movement of the vehicle is restricted. Secondly, path replanning is combined with a multi-path selection strategy, and a dynamic programming algorithm is used to generate the initial path under normal circumstances. ,in is the cost of each node on the path, and m is the total number of nodes; multiple alternative paths are selected based on the preliminary path, and the real-time cost of each alternative path is calculated; finally, the optimal path is selected based on the real-time evaluation, so that the vehicle can quickly bypass obstacles or congested areas, reduce transportation delays, improve the reliability and flexibility of the entire coal mine transportation method, and ensure the continuity and efficiency of production.

[0034] In order to better illustrate the dispatching control of the invention, the following table shows the positioning accuracy of different sensors of the coal mine vehicle dispatching control system. Through the multi-modal sensor integration data and adjusting the weight coefficient according to the noise level, the experiment shows that UWB has the highest accuracy in the coal mine transportation channel. After the data is fused through the Kalman filter algorithm, the positioning accuracy is greatly improved, which can meet the high-precision dispatching requirements.

[0035] Table 1: Positioning accuracy of different sensors in coal mine vehicle dispatching control system

[0036]

[0037] Table 2 shows the results of the fault isolation experiment. The present invention can complete the fault response within 2.5 seconds, and 95% of the faults can be successfully isolated. After the fault is isolated, the average time for path replanning is controlled within 40 seconds, ensuring that the dispatching system can recover quickly after the fault.

[0038] Table 2: Fault isolation experiment results

[0039]

[0040] Table 3 shows the optimization effect of the improved reinforcement learning algorithm. After the optimization of the reinforcement learning algorithm, the task completion time is reduced by an average of 15.2%, which is specifically reflected in the shortening of the completion time in different experimental tasks to varying degrees. The experimental data show that through optimization, the reaction speed of the vehicle in the scheduling process is significantly improved, and the processing efficiency of the system has been significantly improved whether facing routine tasks or emergencies. In particular, in Experiments 2 and 3, the task completion time was reduced by 53 seconds and 52 seconds respectively, which proves that the reinforcement learning algorithm can effectively analyze historical data and dynamically adjust the scheduling strategy during the task execution. In addition, the efficiency improvement is not completely consistent, but is related to the complexity of different tasks and environmental conditions. The experimental results show that the algorithm has strong adaptability in practical applications. In general, the reinforcement learning algorithm not only improves the overall scheduling efficiency of the system, but also verifies its stability and robustness under different conditions, providing a more intelligent solution for the scheduling system.

[0041] Table 3: Optimization effect of improved reinforcement learning algorithm

[0042]

[0043] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for intelligent dispatching and controlling coal mine vehicles based on position feedback, characterized in that: The following steps are involved: S1. First, the location information of coal mine vehicles is collected through a variety of positioning technologies, and the improved Kalman filter algorithm is used to achieve high-precision position feedback control; S2. Secondly, based on the improved reinforcement learning algorithm, the vehicle driving path and task allocation strategy are adjusted dynamically in real time to ensure the efficient execution of the scheduling task; S3, then issue dispatch instructions in real time through the wireless communication network, and monitor and optimize the execution of the vehicle through position feedback; S4. Finally, vehicle faults and emergencies are detected through vehicle status monitoring and safety strategies, and emergency control is automatically performed, including fault isolation and path replanning; The implementation of the emergency control in step S4 includes: S41, firstly, use intelligent control logic to achieve fault isolation, determine whether there is a fault based on historical data and fault threshold settings, and enable safety mode and restrict vehicle movement when a fault occurs; S42, Secondly, for path replanning combined with multi-path selection strategy, under normal circumstances, the dynamic programming algorithm is used to generate the initial path ,in is the cost of each node on the path, and m is the total number of nodes; multiple alternative paths are selected based on the preliminary path, and the real-time cost of each alternative path is calculated; finally, the optimal path is selected based on the real-time evaluation.

2. The method for intelligent dispatching and controlling coal mine vehicles based on position feedback according to claim 1 is characterized in that: The implementation steps of step S1 are: S11, first use GPS, INS, UWB, RFID multi-modal sensors to collect the initial position information of the mine car, and initialize the state , the initial covariance matrix and the measurement noise covariance matrix for each sensor and the process noise covariance matrix ; S12. Secondly, based on the real-time performance data of the sensor, an adaptive noise estimation method is used to estimate the measurement noise covariance matrix of each sensor. , where k is the time, for the observation data of sensor i , update the noise level by measuring the error: ,in is the smoothing coefficient, is the residual, where is the observation matrix, is the state estimate vector; S13. Then, the data fusion of each sensor is dynamically weighted. The weight is proportional to the inverse of the noise covariance. The smaller the noise of the sensor, the greater its weight, and vice versa. The weight of each sensor is calculated as ,in To measure the trace of the noise covariance matrix, the weights are normalized to ensure that the sum of the weights of all sensors is 1; S14, then use the weighted observation value of each sensor To correct the state prediction ,in is the Kalman gain; S15. Finally, use the new weights and Kalman gain to update the state covariance matrix , where I is the identity matrix.

3. The method for intelligent dispatching and controlling coal mine vehicles based on position feedback according to claim 1 is characterized in that: The operation of using improved reinforcement learning to achieve real-time dynamic adjustment of the vehicle's driving path and task allocation strategy in step S2 is: S21. First, define the state space S, and define each state as the vehicle's current location information, current task, and surrounding environment information. , define the action space ; S22. Next, a reward function is designed to evaluate the benefits of each state-action pair, taking into account factors such as task completion rate, safety of the driving path, and minimum energy consumption. The reward function is ,in are weight coefficients, which represent the importance of different factors in the reward function; S23, then use the improved method to update the Q value, the specific formula is: ,in is the learning rate, is the discount factor, is the updated Q value, is the Q value, indicating that in the state Take action The estimated long-term return of For the status The maximum Q value of all possible actions represents the maximum future reward that can be obtained in the next step; S24, then adjust the weight coefficient in the reward function through the adaptive mechanism , dynamically optimize according to the actual situation of task execution ,in It is the result of feedback adjustment based on the difference between the current task execution efficiency and the expected value. is the adjustment rate; S25. Finally, combined with model predictive control, the optimal state of the next step is predicted through the model of the current state to improve efficiency , where MPC is model predictive control.

4. The method for intelligent dispatching and controlling coal mine vehicles based on position feedback according to claim 1 is characterized in that: The implementation method of issuing dispatch instructions in real time and monitoring and optimizing vehicle execution status through position feedback in step S3 is as follows: first, a low-power wide area network architecture is established, and the dispatch center generates dispatch instructions according to the current status of the vehicle and the task priority; then the dispatch instructions are sent to each vehicle through the wireless network, and the message queue is used to ensure the timeliness and reliability of the instructions; then each vehicle regularly collects its own position information and sends it back to the dispatch center through the wireless network, and the vehicle position information is uploaded in real time to form a closed-loop monitoring; finally, the dispatch center receives the vehicle's position information and compares it with the predetermined path, evaluates the execution status and calculates the optimization index; if deviation delay is found, the dispatch center adjusts the dispatch instruction in real time and reissues it.

Citation Information

Patent Citations

  • Machine learning strategy based distance preferred optimal path selection method

    CN109947098A

  • Robot path planning method and device, electronic equipment and computer readable medium

    CN116909293A

  • Goods van route automatic planning system based on GIS (Geographic Information System)

    CN117053819A

  • Emergency ambulance path recommendation method and system and storage medium

    CN118010058A

  • Strip mine unmanned truck dynamic scheduling system and method based on reinforcement learning

    CN118674208A

Cited By

  • Energy-saving type rescue path dynamic planning system and method for intelligent networked automobile

    CN120194723A

  • Intelligent dispatching management system of coal mine monorail crane locomotive

    CN120317597A