Cooperative Localization And Navigation Planning Of Multi-mobile Robot

Mobile robot,which can help human beings to get rid of dangerous and harsh work,has a greater mobility and flexibility than normal robot.At the present stage,mobile robot researchers often tend to focus on single-robot.Whereas,multi-robot is superior than single-robot in sensing range,computing power and robustness,thus it has a promising prospect in some applications.Self-localization and navigation-planning are the bases of mobile robot to complete some complex tasks.This dissertation mainly focuses on cooperative-localization and navigationplanning of multi-robot.We have achieved the hunting of an unpredictably moving target by a group of robots,which is verificated on simulation platform and our laboratory-designed experimental platform.Besides,visual target recognition and initial localization are also included.Firstly,we designed a visual target recognition algorithm based on Bayes' Theorem.In the experimental platform,we use LEDs with different colors to characterize different robots,and assume that the distribution of the LEDs' colors in the RGB space is Gaussian.Through the preceding collected samples,we realized the recognition of different LEDs of the robots.Secondly,during the initial stage of the robots,a simple and efficient static initial localization algorithm has been proposed.Aimed at optimize the raw localization data caculated by triangular-geometric method,we collecte redundant sensor data on the robots,and construct a nonlinear least squares optimization problem.Gauss-Newton method or LevenbergMarquardt method was utilized to solve the optimization problem.Thirdly,we learn from the experience of wireless sensor network,including distributed measurement,centralized estimation and real-time wireless communication.According to the specific motion model and observation model,performance and limitation of sensors on the robots,we developed a dynamic cooperative localization method which is based on Extended Kalman Filter algorithm,and achieved the dynamic cooperative localization of multi-robot and target tracking during motion stage.Finally,based on the methods mentioned above,e.g.,the visual target recognition,static initial localization,and dynamic cooperative localization,we developed a dynamic navigationplanning method.Then we realized formation planning,trajectory planning,velocity planning,and velocity control,respectively.In trajectory planning,in order to obtain a trajectory with lower curvature,sequence quadratic programming is utilized to optimize the planning trajectory based on the third-order Bézier curve.In velocity planning,we use model predictive method to achieve smooth planning velocity of the robots.Eventually,on the experimental platform,we successfully achieved the hunting of an unpredictably moving target by multi-robot in an indoor environment with boundary constraints.