A Research On Multi-robot Cooperative Localization By Fusing Multi-sensor Data

Localization is a key capability for both individual robots and for robot teams.For robot teams,however,one must distinguish between two kinds of localization:absolute localization capability and relative localization capability.There's many benefits in cooperation between multiple robots.Among other benefits,cooperation can improve their localization capabilities.This paper mainly disguss the autonomy robots fuse proprioceptive data and exteroceptive data based on different algorithm.The robots would be the beacons each other,so as to improve the precision of localization.The colony system framework of the robots and the point models of mobile robots based on variance wheel are discussed in chapter 2. The numerical relationship of noise models is analysed and the observing models is also introduced.This chapter is the base of this paper.A data fusing algorithm based on Extended Kalman Filter(EKF) is presented in chapter 3.The simulation results show that this algorithm can accord with the request of precision.But the disadvantage of this algorithm is hign precision request for the initial condition.The data fusing algorithm based on Particle Filter(PF) and another improved Particle Filter is discussed in chapter 4. The simulation results show that these algorithms have the same level on presicion with EKF and less request for initial condition.These algorithms also are more robust than EKF ,but computationally expensive.In a next place,the algorithm of PF-EKF is applied on the situation of no preparative information or big error in the initial condition.In this cases,PF is used to converge the initial condition to satisfy EKF.Then,EKF is used to be iterative filter. The simulation results show that this algorithm can solve the contradition between the operative precision and the operative magnitude efficiently.In practice,if there are many units in the robot team,cooperative localization between all of the robots will make the operation to be inefficient and make the precision down.To solve this problem,a target-choosing method based on the maximum of information entropy is presented in chapter 5.Based on the theory of information entropy,this algorithm combines EKF to choose the referenced targets. The simulation results show that this algorithm decreases the cost of computation distinctly and contents the request for precision.The last section sums up the contant of the thesis and point out the questions which should be resolved in the future.