A framework for control of formations of mobile robots: Theory and experiments
We describe a framework for coordination and control of a group of nonholonomic mobile robots that allows us to build complex systems from simple controllers and estimators. The resultant modular approach is attractive because of the potential for re-usability. Our approach to composition also guarantees stability and convergence in a wide range of tasks. There are two key features in our approach: (1) a paradigm for switching between simple decentralized controllers that allows for changes in formation, and (2) the use of information from a single vision based sensor—an omnidirectional camera, for all our controllers. We utilize estimators that abstract the sensory information at different levels, enabling both decentralized and centralized cooperative control.
We address the problem of stabilizing a group of mobile robots in formation. The group is required to follow a prescribed trajectory, while achieving and maintaining a desired formation shape. Each individual robot is equipped with several control algorithms that allow the robot to control its position and orientation with respect to neighboring robots or obstacles in the environment. By using a decentralized leader-following approach, follower robots are able to automatically switch between continuous-state control laws to achieve a desired formation shape. The stability properties of the closed-loop switched system are studied using Lyapunov theory. We describe algorithms for assigning control policies to different robots, based on sensor and actuator constraints. This assignment is described by a control graph. We relate the structure of the control graph to the stability of the dynamics of the formation.
Our framework allows us to dynamically reconfigure a team of autonomous robots. Reconfiguration means that a team of vehicles is capable of reaching a desired formation (a goal position and a specified geometric shape) while preserving the stability of the entire system. We describe the dynamic reconfiguration problem as a graph-based assignment problem, and develop an approach that allows decoupling the problem of formation shape assignment from that of control graph assignment and accomplish both in a decentralized fashion. We accommodate obstacle avoidance in dynamic environments using a virtual leader concept, without modifying the overall formation control framework. We examine both holonomic and nonholonomic mobile robot formations, presenting analytical stability results and numerical simulations illustrating our approach. Finally, to verify the real world applicability of our framework, we present experimental results with a team of car-like robots equipped with omnidirectional cameras and wireless network cards.