Underwater robot orbits targets despite losing visual contact
AquaOrbit: Sim-to-Real Reinforcement Learning for Underwater Target Orbiting under Intermittent Visual Feedback
Robotics
Summary
Losing sight of a moving target underwater makes it hard for robots to keep circling around it and find it again. The authors developed AquaOrbit, a controller that helps underwater robots continue orbiting even when vision cuts out temporarily by using past information about the target’s position and robot’s motion. They trained this system entirely in simulation and showed it works well on real robots without extra training, handling different target movements and vision interruptions. Their system outperforms traditional control methods, especially when vision is lost for several seconds.
What this means in practice
- •For underwater robot operators: Improve underwater robot target tracking and maneuvering under challenging visual conditions using learned recovery strategies for orbiting tasks.
- •For marine environmental monitoring teams: Enable robust monitoring of moving underwater objects despite temporary visual obstructions using AquaOrbit’s sim-to-real trained controller.
Authors
Kanzhong Yao, Jinyi Leng, Hao Zhang, Zhe Sun, Xuelong Li
Abstract
Intermittent visual loss disrupts target-relative feedback during underwater orbiting, making it difficult to maintain coordinated motion and reacquire a moving target. We present AquaOrbit, a reinforcement-learning controller with a recovery module for underwater target orbiting under interrupted visual feedback. During detection loss, the recovery module uses latched line-of-sight, roll, and depth references to support stabilization and target reacquisition. We train the controller in Isaac Sim with dynamics, observation, and vision-loss randomization. Evaluated without retraining in Gazebo/ROS2 under a different physics engine and perception perturbations, AquaOrbit completes 20/20 orbiting trials in each of the static- and moving-target conditions on an unseen variable-depth 3-D trajectory. In the moving-target condition, it reduces mean line-of-sight error by approximately 46% relative to a PID-based visual servoing controller with recovery while maintaining comparable path-tracking accuracy; removing the recovery module reduces completion to 9/20. Zero-shot physical deployment with fully onboard perception and control demonstrates elliptical, figure-eight, and variable-depth circular trajectories, including the latter two trajectory types absent from training. The robot maintains attitude stability during manual occlusions lasting up to 8s and reacquires the target within 2.5s in the reported attitude-induced field-of-view loss events.