Rotorcraft guidance with sampling based model predictive control

A key technology in automation of rotorcraft flight is collision-free guidance. Especially in operations close to the ground, detection and automatic avoidance of ground-based obstacles and aircraft is a demanding task. This paper presents a sampling-based model predictive approach for collision-free guidance of rotorcraft. The method calculates control inputs for the flight controller by predicting the closed-loop dynamics of the rotorcraft for a short time horizon and evaluating the predictions with a cost function, which can take an arbitrary form. The approach implements a simple algorithm, mitigating the need for iterative optimization and allowing for deterministic execution time. The cost function is set to ensure collision-free maneuvering while following a desired path, as well as considering constraints of the rotorcraft states and control inputs. The path following performance is tested in closed-loop simulations with a non-linear helicopter model. The algorithm is implemented on a graphics processing unit for parallel execution, strongly decreasing the computation time.