Coevolutionary model predictive formation control and localization of swarm robotic systems

Abstract

Coordination of swarm robotic system is a promising alternative to a single robot system because it provides a higher level of robustness as a result of its redundancy and the potential for simpler functionality in each robot. Moreover, the possibility of conducting work in parallel allows various applications, e.g., cooperative transport, reconnaissance, coverage, and exploration tasks. In order to perform these tasks, robots must be capable of controlling their formation stably while performing self-localization. This thesis therefore focuses on three research issues in swarm robotic systems: distributed formation stabilization, collision avoidance, and localization. In the first part, this thesis proposes cooperative coevolutionary algorithm (CCEA)-based model predictive control (MPC) that guarantees asymptotic stability of swarm robot formation. While conventional evolutionary algorithm (EA)-based MPC approaches cannot guarantee stability, the proposed CCEA-based MPC approach guarantees asymptotic stability regardless of the optimality of the solution that the CCEA-based algorithm generates with a small number of individuals. To guarantee stability, a terminal state constraint is found, and then a novel repair algorithm is applied to all candidate solutions to meet the constraint. Furthermore, as the proposed CCEA-based algorithm finds Nash-equilibrium state in a distributed way, robots can quickly move into a desired formation from their locations. Numerical simulations and actual experiments demonstrate that the CCEA-based MPC greatly improves the performance compared to conventional particle swarm optimization (PSO)-based MPC. In the second part, this thesis proposes a novel receding horizon particle swarm optimization (RHPSO)-based MPC for swarm robot formation incorporating collision avoidance and control input minimization. Formation control problem incorporating collision avoidance and control input minimization can be formulated as a constrained n...