Energy-saving profile optimization for underwater glider sampling: The soft actor critic method

Abstract

Underwater glider is developed for low-cost oceanic exploration. This work aims to solve the energy optimal yo-yo numbers for underwater glider sampling profiles. Firstly, the motion simulator is developed. By specifying the target sampling paths, the navigational constraint is designed to guide the trajectory and to evaluate the sampling quality. The energy consumption of different numbers of yo-yos is analyzed. The problem framework of profile optimization is developed. To solve the proposed problem, a soft actor critic based method is proposed. The proposed method avoids repetitively optimizing by normalizing the remaining progress of sampling into standard inferring inputs. Through simulations, the results of soft actor critic profile optimization regarding energy consumption and sampling performance are comparable to those of meta-heuristic algorithms and Lipschitz continuity based optimizations. Moreover, the proposed method shows the computational advantages. This work provides a reference to estimate and save the energy consumption of underwater gliders.