Hybrid Model for Trajectory Planning of Agile Autonomous Vehicles

This paper presents a hybrid architecture for autonomous trajectory planning of agile vehicles. A velocity control system providing the ability to accurately track trajectories is combined with a maneuver scheduler that enables execution of pre-programmed agile maneuvers. The closed-loop dynamics under this control architecture are described by a simple hybrid model, consisting of a set of constrained, linear time-invariant modes and discrete fixed-duration transitions in the state space. Given these dynamics, mixed integer linear programming (MILP) is used to compute optimal trajectories in cluttered environments. Continuous constraints and binary logic are combined to model the constraints governing the dynamics, to formulate switching rules between different velocity modes, to encode execution of agile maneuvers, and to account for obstacle avoidance. Both offline time-optimal planning and online receding horizon formulations are presented. The framework is applied in detail to a small-scale helicopter, for which several receding horizon results are given. A discussion about practical considerations regarding a real-time implementation concludes the paper.