Robust Control of Coupled Multi-robot Systems under Signal Temporal Logic Specifications with Collision Avoidance

Abstract: Multi-agent systems have been gaining attention among robotics researchers due to their capabilities, in terms of which they surpass individual robots. Such systems, however, entail an increase in complexity, especially to the task planning problem. Therefore, a more sophisticated planning approach is needed, one that is able to resolve conflicts between local tasks, in particular. These conflicts arise from couplings between robots expressed in their task specifications. A computationally inexpensive collaborative control scheme is implemented in a setup consisting of three omni-directional robots. The robots execute tasks encoded in Signal Temporal Logic, using Prescribed Performance Control. They are guaranteed to satisfy the specifications or, if that is not possible, reach a least violating solution, even in the face of a task conflict. What is more, collision-free trajectories are produced. A series of simulations and experiments is carried out to verify the utility of the method. The approach is shown to be effective despite restrictive assumptions imposed on the system. Finally, directions for future work, that would allow the method to perform well in more demanding experimental scenarios, are suggested.
 

Examiner: Prof. Dimos Dimarogonas