Bearing-only Observations Multi-Robot Formation Control

Multi-robot formation control(MRFC)involves many disciplines such as mechanicals,electronics,computers,control,communication and artificial intelligence.It has been widely used in national defense,industry,agriculture,everyday life and other fields.Formation control,as a fundamental problem in the research of multi-robot,is the basic requirement for multi-robot to accomplish complex tasks and has received immense attention from both academia and industry.There has been a lot of research on the MRFC based on the distance and/or position observation.However,this kind of formation control method needs to deal with a large amount of information and the algorithm is complex.In addition,it is difficult to obtain the distance or position observation under certain special circumstances,which limits the practical application of the method.MRFC with bearing-only information overcomes the limitation of MRFC that based on the distance and/or position observation.It has been become a new research hot point in MRFC.Focusing on the controllability,observability and stability of a class of Leader-Follower(LF)MRFC algorithm and based on the rank theory for nonlinear system and the real-time panel method,we have studied several key issues involved in the LF-MRFC problems such as state estimation,closed-loop control,trajectory tracking,robot localization and autonomous navigation,among others.The contribution of this thesis can be classified into three parts:1.For the(cascade)LF-MRFC system based on bearing-only observation,it has been proven that the bearing-only observation fulfils the controllability and observability condition for general nonlinear LF system.To deal with the noises in the observation of the Follower robot,the unscented kalman filter has been employed for improving its observation accuracy.A closed loop control system using input-output state feedback is designed to guide the Follower robot to accurately track the Leader robot.2.In unstructured environments,it is rare for the robot to precisely follow the scheduled trajectroy which has been a key challenge to the MRFC.To solve the problem,a particle filter has been proposed for the Leader robot localization based on bearing-only observation of landmarks in the bearing-only LF-MRFC and cascade LF-MRFC system.The observability of the system has been analyzed,and the deviations between the true trajectories and the planning trajectories of the Leader robot can effectively be reduced using the proposed approach,improving the MRFC performance in unstructured environments.3.In realistic environments,obstacles are inevitable when executing the MRFC task.Then,the robots should not only keep the formation,but also reasonably avoid the obstacles.To this end,an autonomous navigation and obstacle-avoidance approach has been proposed based on the panel method,which enjoys a high computational efficiency since the calculation is based on the bounds of the obstacles rather on all the surface.This approach has also been implemented on the bearing-only LF-MRFC system.For the sake of evaluating the effectiveness and validity of the proposed approaches,a large number of tests have been carried out both on the synthetic data under the simulation environments based on Matlab and Webots softwares,and on the real world data based on the Turtlebot robot experiment platform.Both simulation and experiment results show the real time performance and effectiveness of the method/algorithm.