Multi-robot formation control

: a receding-horizon leader-follower framework

Abstract

Over the past decades, the coordinated controls of multi-robot systems have received considerable attention due to the fact that multiple robots are envisaged to help accomplish tasks not possible with individual robots acting alone. Formation control, which is the most important research area in multi-robot coordination, aims in controlling the relative positions and orientations of robots in a group while allowing the group to move as a whole. To successfully execute various tasks in a dynamic and uncertain environment, robots engaged in multi-robot tasks must alter their formation according to the changing environment, and in response to unexpected changes in the task requirements. Examples of formation control tasks include: assignment of feasible formations, moving into formations, maintenance of different formation shapes, and switching between formations. This thesis aims to develop a new leader-follower control framework to control a group of non-holonomic mobile robots in formation tasks, and furthermore, to incorporate the receding-horizon method into the leader-follower strategy to achieve fast convergence rate for the whole formation control framework. The research is carried out mainly in the following perspectives. First, a new leader-follower control framework with the consideration of separation, bearing and orientation derivation between leaders and followers together, is developed. With an assumption that all separation, bearing and orientation deviation are enumerated in advance, two control schemes, Separation-Bearing-Orientation Scheme (SBOS) and Separation-Separation-Orientation Scheme (SSOS), are proposed for two-robot and three-robot formations, respectively. The major difference of the proposed schemes from the literature works is that the orientation is explicitly controlled, which helps achieve more effective formation activities of a multi-robot group. For example, many existing leader-follower methods could not address the formation problem when the robots move backwards together, while the problem can be solved by the approaches reported in this thesis. Extending the proposed leader-follow approach, a Separation-Obstacle-Orientation Scheme (SOOS) can be further developed for leader-obstacle problems. Second, the receding-horizon (RH) strategy is incorporated to the proposed leader-follower approach to improve the convergence rate of the defined formation tracking errors. A rapid converging speed of robot state is critical since it directly affects the overall formation performance of the whole group. Through proper design of the RH procedure, a novel receding-horizon leader-follower (RH-LF) control scheme is developed to yield exponential convergence of the robot formation tracking errors in the terminal state region. Note that most formation controls in existing literature were investigated based on stability analysis only, and few of them were based on stability as well as performance optimization to achieve fast convergence. Third, solid experimental studies have been performed on a group of multiple mobile robots in the laboratory. Experimental results demonstrate that the proposed new leader-follow strategy can effectively control the robots in formation tasks including backwards formation. By adding a receding-horizon strategy to the leader-follow formation control scheme, the converging performance of the formation errors is improved greatly.

Date of Award	15 Jul 2009
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Dong SUN (Supervisor)