Energy-Optimal Model Predictive Control for Unmanned Underwater Vehicles in Offshore Aquaculture Fish Net-Pen Visual Inspection

Unmanned underwater vehicles are deployed to automate the production processes in offshore aquaculture, but the onboard power supply with limited energy capacity constrains the operational range and time. In this paper, a nonlinear energy-optimal Model Predictive Control (EO-MPC) is proposed to perform a 4-degree-offreedom 3D fish net-pen visual inspection trajectory tracking while minimizing energy consumption. The EOMPC problem with explicit energy-related terms in the performance index (PI) is transcribed into a nonlinear programming problem (NLP), solved via IPOPT, the open-sourced primal-dual interior point solver. Using the specifications of Blue Endeavour Project (the upcoming first-of-its-kind offshore salmon firm in New Zealand) of the New Zealand King Salmon Company and the work-class ROV called RexROV 2, theoretical fundamentals and practical implementation aspects are detailed, and four controllers are tested in high-fidelity simulation using Robot Operating System and Gazebo Physics Engine. In a general constrained operational working environment, the proposed EO-MPC controller saves 3.1% - 21.4% more energy than the conventional MPC (CO-MPC) while achieving better or equivalent trajectory tracking performance under different underwater current disturbance speeds (0.0m∕s, 0.5m∕s and 0.9m∕s).