Research On Navigation System Of Coal Mine Rescue Robot

Coal is one of the major energy sources in China,while coal mining is a high-risk industry.It is critical to improve the efficiency of coal mine disaster rescue and ensure the safety of the rescue process.Coal mine rescue robots can replace the rescuers and go into the mine to detect the scene of the disaster situation,which can improve the efficiency of rescue process and ensure the safety of rescuers.The navigation function can make the robot construct the destroyed map after the accident,provide important reference for the rescue work,and assist the robot in the path planning and autonomous or remote control,which can improve the rescue efficiency significantly.The development of navigation system is of great significance to improve the autonomous performance and intelligence of coal mine rescue robots and improve the efficiency and practicability of rescue task.In this paper,we use the CUMT4 coal mine rescue robot as a platform to build suitable navigation system for the application in coal mine.In order to meet the needs of coal mine rescue robots,the overall design of the navigation system is carried out.The hardware system is divided into five parts:control system,communication system,sensing system,power supply system,power and drive system.For each part,The hardware is selected and the circuit connection is designed.The characteristics of ROS operating system is analyzed on the basis of comparing the existing software system,.The software system is divided into six parts:environment perception system,coordinate transformation system,chassis drive system,SLAM system,path planning system and human-computer interaction system.The overall function of the various parts are designed in this part.For the question of simultaneous localization and mapping of coal mine rescue robots,several SLAM algorithms are researched,designed and selected for coal mine environment.The technical difficulties of SLAM technology are analyzed.The SLAM problem is described by using the probability model and the optimization model.Based on this model,the mathematical principles of online SLAM and complete SLAM algorithm are introduced.Firstly,EKF-SLAM of extended Kalman filter based on Bayesian recursive method,and Monte Carlo localization and Gmapping algorithm based on particle filter are studied.Simulate the EKF-SLAM process by matlab and the results shows that the correction of the observed model has obvious correction effect on the prediction of the motion model.Monte Carlo localization simulation test shows that use the motion equation as a recommended distribution can get a good localization performance when the map is known.Then,the principle of PLICP scanning matching is introduced.Based on the Gmpping algorithm,the PLICP-FastSLAM is constructed by using the information of localization of PLICP instead the odometer model in the proposed distribution function,which has a better adaptability in the case of the odometer can't work well.The principle of Hector-SLAM scanning matching algorithm is illustrated.Finally,the process of Cartographer-SLAM algorithm is analyzed,and the optimization of the back-end diagram is described.In order to compare the practicality of the three algorithms,test in the indoor environment.First the localization effect of PLICP algorithm is verified.Then the accuracy of the map built by PLICP-FastSLAM,Hector-SLAM,Cartographer-SLAM algorithm mapping is compared in the same corridor environment.The results are analyzed and the realization processes are compared,as well as advantages and disadvantages of each algorithm.The experiment results show that PLICP-FastSLAM uses the improved particle filter through the resampling process,and uses the odometer and scanning matching to reduce the number of particles,which greatly improves the efficiency and precision of the map building,and has good adaptability to the indoor environment.Hector-SLAM has the highest efficiency in scan matching.The map has high accuracy but poor adaptability for large rotations.Hector-SLAM dependents on high-scanning frequency of the laser sensor.Cartographer-SLAM algorithm is based on graph optimization and use the sparse pose adjustment,which can build the highest consistency of the map,and has little affected by environmental noise.Three algorithms can all get good quality maps in the indoor environment.Aiming at the path planning problem of coal mine rescue robots,the global path planning algorithm based on A*and the local path planning algorithm based on DWA are studied.This paper analyzes the characteristics of A*algorithm,BFS algorithm and Dijkstra algorithm,and analyzes the realization process of A*algorithm.The DWA algorithm is described in detail.In order to integrate the two algorithms to the CUMT4 navigation system,the ROS-based path planning system architecture is introduced,and the cost map,path planning algorithm,location information,sensor information and coordinate transformation relation of navigation stack are analyzed.Finally,the simulation test of indoor map generated by Cartographer-SLAM shows that the ROS-based path planning system has the advantages of fast calculation,short and smooth path and so on,which can meet the navigation using.Aiming at the concrete design and implementation of the navigation system of coal mine rescue robots,the hardware and software of CUMT4 robot navigation system are designed in detail.The laser,depth camera,AHRS,odometer and other underlying drive nodes are designed.The underlying controller hardware is designed by the use of STM32 control core,then the software implementation process is introduced.Finally,the navigation system using the various systems are combined to achieve the navigation function.In order to verify the practicality of the CUMT4 navigation system,scene tests are carried out in simulated mines.In this paper,we compare the use of PLICP-FastSLAM,Hector-SLAM and Cartographer-SLAM algorithms in field test,especially comparing the mapping accuracy and calculation pressure of the three algorithms.The results show that Cartographer-SLAM has the best robustness,highest accuracy and consistency,and the minimum calculating pressure.Finally we use the map constructed by Cartographer-SLAM to carry out autonomous and semi-autonomous navigation tests.The results show that the CUMT4 navigation system has a good effect and can basically meet the use of coal mine environment and achieve the desired effect.