Gas Source Localization Using Multi-robot Based On Conditional Information Entropy

Research on gas source localization technology(or active olfactory technology)based on mobile robots began in the 1990 s.With the continuous advancement of science and technology,mobile robot technology and gas sensor technology have also been developed,and the active sense of smell has evolved from a single robot working mode to a multi-robot collaborative mode,which can complete more and more complex tasks and greatly improve work efficiency.This paper mainly studies the multi-robot gas source localization based on conditional information entropy.The distributed multi-robot structure and the information gradient search(i.e.information tropism)strategy are used to estimate the gas source location.The specific work content is as follows:Firstly,the mobile robot experimental platform is built and the overall design and layout of the modular and structured of the software and the hardware system are completed according to the actual needs,including: selection and configuration of various sensors,hardware layout,design of the underlying motion control system and the main program of the upper layer,the construction of the adaptive token ring network.Secondly,aiming at the linear lag of metal oxide gas sensor,a method to improve the gas velocity is proposed.The first-order linear inertia delay mathematical model of metal oxide gas sensor is established.The two-sensor M-sequence experiment and ? model method are used to find the practical time constant of the gas sensor is derived,and the reconstructed value is derived from the measured value and used instead of the measured value.The method verifies its effectiveness through experiments and lays a good foundation for improving the working efficiency of multi-robot gas source positioning.Thirdly,this paper proposes a gradient search method based on conditional information entropy and establishes a corresponding mathematical model.This method is a distributed algorithm,which estimates the position of the gas source through distributed particle filter,and drives the motion of the robot by conditional information entropy gradient between the position of the gas source and the measured value.A single robot uses distributed particle filter to estimate the position of gas source,and calculates the conditional information entropy,and then obtains the gradient of conditional information entropy.In the positioning process,the gas concentration is measured and binarized by a single robot.The distribution of gas source location is obtained by odor bag backstepping method,and the distribution function parameters and their own location information are distributed.At the same time,the distribution function parameters and location information issued by other robots are received in one control cycle.According to the information obtained,the distribution function is filtered and simplified.The joint distribution is used to update the particles,and the estimated gas source location and conditional information entropy are calculated.When the particle convergence radius and conditional information entropy value are lower than the set threshold for a long time,the algorithm stops and the final estimated gas source location is obtained.The effectiveness and superiority of the proposed method are demonstrated by experiments in a flat barrier-free environment and comparison with gas source localization experiments of single-robot.Finally,this paper summarizes and prospects the multi-robot gas source localization method based on conditional information entropy.The purpose of this paper is to provide a way to study the multi-robot gas source positioning in the later stage.