This talk covers recent developments in the use of Mixed-Integer Linear Programming (MILP) and Model Predictive Control (MPC) with constraints. The speaker's research group are using MILP to capture non-convexity in trajectory optimizations for autonomous vehicles, particularly Unmanned Aerial Vehicles (UAVs) and spacecraft. MILP is used here in a new MPC formulation, shown to offer finite-time maneuver completion. It improves on existing formulations by enabling the use of general cost functions and target states. Further modifications are shown to guarantee robust feasibility in the presence of a bounded disturbance. This approach is equivalent to planning conservatively for the far future, leaving `room' for future feedback action.
The talk begins with an overview of the work of the speaker's research group, including projects, target scenarios and testbeds. The MPC formulations are presented with illustrative simulation examples. Finally, results from recent hardware experiments are shown.
Back to Control Seminars Page