A dual-mode model predictive controller for robot formations

Control of nonholonomic autonomous robots by using an input-output feedback linearization technique has been well explored. It has been noted that model predictive control (MPC) methods may have advantages over state feedback laws when applied to mobile robots, including consideration of constraints on inputs or state vectors. However, MPC algorithms require on-line optimization, resulting in a significant computational burden for large systems or formations of robots. In this paper, we develop a dual-model MPC algorithm which uses an input-output feedback linearization controller for states-within a specified terminal constraint set. Control of formations of nonholonomic robots in leader-follower configurations is simulated off-line, and performance characteristics of the dual-mode MPC controller are contrasted with those of the input-output feedback linearization controller alone.