Moving Horizon Planning and Control Under Uncertainties with Guarantees – Combining Operational Choices and Motion Primitives

We present a method for enhancing autonomous system capabilities by optimizing both planning and control layers over a moving horizon. Traditional reliance on simplified models is overcome by reformulating the planning problem into a robust optimization issue and leveraging multiple operational choices. Constraint satisfaction and collision avoidance are assured through a technique called constraint back off, which adapts to the system's operational mode, derived from precise motion primitives. The path planning problem is then converted into a solvable mixed-integer linear optimization problem. Using existing robust model predictive control strategies for the control layer allows for handling uncertainties and rapid response to changing obstacles. Simulation results validate the efficacy of our approach.