Energy-Optimized Hybrid Collision Avoidance for ASVs

This paper considers the development of a hybrid planning and collision avoidance architecture for autonomous surface vehicles (ASVs). The proposed architecture combines a high-level optimization-based planning algorithm with a mid-level collision avoidance (COLAV) algorithm based on model-predictive control (MPC). The high-level planner produces an energy-optimized trajectory by solving an optimal control problem via a pseudospectral method, taking into account known static obstacles and ocean currents. The mid-level algorithm performs MPC by solving a nonlinear program (NLP) to produce a collision-free local trajectory, also taking into account dynamic obstacles. In particular, the NLP optimizes for a combination of following the energy-optimized trajectory with performing readily observable maneuvers, as defined by Rule 8 of the International Regulations for Preventing Collisions at Sea (COLREGs). Numerical simulations are used to verify that the hybrid architecture produces safe, efficient and readily observable trajectories.