Using Information Entropy To Search And Localize A Gas Source In A Cluttered Scene

The leakage of flammable,explosive and toxic gases threatens people’s lives and property.It is of great significance in industry and life to detect gas leakage and then find the gas sources in time before the disaster expands further.In order to solve the research difficulties brought by gas propagation characteristics and consider the ubiquity of obstacles in the actual sourcing scene,this research uses a mobile robot as the sourcing platform to search and localize a gas source based on information entropy in a cluttered scene.After the realization of the search task,gas source-related parameters such as gas release rate,lifetime of the emitted material,and gas diffusion rate can be simultaneously estimated based on the probability density distribution to help rescuers predict the gas diffusion state.Firstly,the software and hardware framework of the mobile robot platform for gas sourcing is built,then a complete sourcing system is planned.The platform is equipped with lidar to realize the localization and mapping for robot using SLAM algorithm.The gas sensor and the integrated sensor of wind direction and wind speed provides data input for the sourcing algorithm.The sourcing algorithm outputs the target pose and converts it into a navigation module to output the robot control command,and then performs motion control on the slave computer based on the Socket communication.In addition,an octree-based repetitive pose detection algorithm is proposed to improve the sourcing objective function and improve the performance of 3D point cloud construction.Secondly,the gas plume detection strategy based on zigzag search is studied.This strategy moves the robot in the upwind direction and maintains a certain lateral movement distance.According to the zigzag trajectory movement,the probability of robot contact with plume can be effectively increased to improve the efficiency of plume detection.As the first step of gas sourcing,plume detection needs to construct a certain range of search areas at the end of the link to provide spatial constraints for plume information tracking steps.Thirdly,the plume information tracking algorithm based on particle filter and information entropy is studied.Particle filter is used to estimate the state of gas source.The information entropy algorithm uses the current state of particles to predict the target posture of the next movement of the robot.Both of them iterate continuously to realize the convergence of particle and estimate the parameters of gas source.Aiming at solving the problem that the robot searches the same region repeatedly,an octree-based repetitive planning point search algorithm is proposed,and the entropy score function is improved using the search results,which improves the efficiency of sourcing in the cluttered scenes.In the process of loss of plume,a planning algorithm for edge target points in the search area is proposed to help the robot turn back to the search area,which improves the robustness of searching.Finally,the gas source confirmation rule is studied.The rule combines the gas concentration information and the gas direction probability information to discriminate the correctness of the gas source location parameters estimated in the plume information tracking process.Considering the existence of obstacles in the process of sourcing,a method of boundary planning for recognition area in cluttered scene is proposed,which can prevent the collision between robot and obstacle without losing the ability of gas source identification.In order to deal with the result of pseudo-source judgment caused by gas accumulation on the windward side of obstacles,a pseudo-source disposal algorithm based on artificial potential field is proposed,which can be used to expel the particles from the recognition area and prevent the particles from convergence again,the algorithm can improve the accuracy of sourcing.