New reinforcement learning method optimizes underwater vehicle power budgets

Researchers have developed a novel approach to control underwater vehicles by integrating energy efficiency directly into the reinforcement learning process. This method formulates control as a constrained Markov decision process, allowing for an explicit power budget to be set in physical units. A PPO-Lagrangian algorithm is used to solve this, with a dual variable updated online to meet the budget for each specific vehicle and task. Testing in the MarineGym simulator demonstrated significant power reductions of 14-65% across various vehicles and tasks compared to a task-accuracy-only baseline, while maintaining smooth operation and task accuracy. AI

IMPACT This research offers a tuning-free method for energy-efficient control of underwater vehicles, potentially extending mission ranges and endurance.

RANK_REASON The cluster contains a research paper detailing a new method for reinforcement learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• MarineGym
• PPO-Lagrangian
• Yinuo Wang

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →