Cooperative Pose Estimation And Target Tracking For Multiple Mobile Robots

The mobile robot system,which plays an important role in the industrial and military fields,is currently a continuing research hotspot.It integrates the research results of machine vision,probability theory,control theory,artificial intelligence and some other disciplines.Robots are equipped with a variety of sensors to sense the environment and extract key information to make appropriate decisions.Compared with the single robot,multiple robots with stronger fault tolerance and more flexible deployment structure can accomplish complex tasks through information interaction and collaboration.In this paper,based on the extended Kalman filter,multi-robot cooperative target tracking and pose estimation are studied.The main contributions are addressed as follows:A mobile multi-robot experimental platform based on wireless sensor networks is built.The robot with a unique identification mark is divided into a sensing layer,a driver layer,a communication layer and an operation center according to the function of equipped modules,which possess the ability of sensing external information and exchanging measurement data.Because of the limitation of the individual's perception ability,the designed robot can't achieve pose estimation or target tracking on its own,and it needs cooperation with other robots.The problem of tracking the maneuvering target based on bearings-only when robots are stationary is described.Due to the nonlinear characteristics of the angle observation,the positioning accuracy is affected by relative positions between the target and the observation station,so the geometric dilution of precision based on the least square positioning is analyzed.For the problem that the current statistical model needs to preset the maneuver frequency which can't be adjusted with the target maneuverability changing,a frequency adaptive algorithm based on residual detection is improved.The simulation results show that the improved algorithm achieves better tracking performance than using only the current statistical model.Based on the actual experimental environment,the estimation accuracy of the Newton-Raphson iteration method,the orientation search method and the two-circle intersection method are compared for pose initialization problem of the robots.For the multi-robot cooperative pose estimation problem,a centralized algorithm and a paired algorithm are proposed,in which the centralized algorithm obtains higher accuracy but consume more computing resources while the paired algorithm gets lower accuracy but makes calculation amount allocated by each robot.Based on multi-robots cooperative pose estimation,for the problem of simultaneously achieving target tracking and dynamic pose estimation,the multi-robots observation scheme is determined,and a low-dimensional filter is proposed to estimate the target state separately.The estimation accuracy of above-mentioned algorithm and the centralized filtering algorithm is compared by simulation.According to hardware conditions of the experimental platform,an appropriate algorithm is selected for the target interception experiment.