Energy-Optimal Control for Autonomous Underwater Vehicles Using Economic Model Predictive Control
Niankai Yang · Dongsik Chang · Matthew Johnson-Roberson · Jing Sun
TL;DR
An economic model predictive controller cuts underwater vehicle energy use during waypoint tracking, nearly matching offline optimal trajectories and beating a line-of-sight baseline in simulation.
BRIEF
Autonomous underwater vehicles can only stay out as long as their batteries allow, so saving control energy extends how far and how long they can work. This work designs a model predictive controller that directly minimizes energy while the vehicle travels between waypoints. It adds a terminal cost that estimates the energy still needed to reach the goal, built from offline optimal trajectories, and it adjusts for ocean currents. In simulations under several flow conditions, it comes close to the offline optimum and beats a line-of-sight baseline.
Abstract
Improving endurance is important for autonomous underwater vehicles (AUVs) as it affects the operational cost and application range of the vehicle. In this article, we propose an economic model predictive control (EMPC)-based controller to reduce the control energy of AUVs while performing waypoint tracking. The proposed EMPC controller optimizes stage costs capturing the control energy consumed within the prediction horizon and a terminal cost approximating the energy-to-go, the energy required to reach the desired waypoint from the end of the prediction horizon. To approximate the energy-to-go, we partition it into the dynamic and static segments based on the operational characteristics of the optimal vehicle maneuver obtained from off-line trajectory optimization using direct collocation (DC). To account for the disturbances caused by ocean currents, we adopt the energy-to-go to a virtual Earth-fixed frame that transforms the drift in the vehicle location to the drift in the desired waypoint. Theoretical and numerical analyses of the approximated energy-to-go reveal that the proposed controller can balance the tradeoffs among energy components spent for vehicle surge, heave, and yaw controls in consideration of vehicle dynamics. Simulations under different flow conditions are conducted to compare the proposed approach with DC and a line-of-sight (LOS) guidance-based approach that optimizes vehicle surge speed for energy minimization. Through simulations, it is shown that the proposed approach achieves near-optimal performance as DC and outperforms the LOS-based approach.