Behavior Trees for Adaptive Multi-sensor Perception in Autonomous Underwater Robotics

Behavior trees (BTs) have recently shown great promise in improving how robotic systems make decisions and execute tasks. Building on such successes, this study explores the potential application of BTs for multi-sensor perception in autonomous underwater robots. We discuss a framework in which BTs could enable subsea robots to adapt to the ever-changing conditions underwater through an intelligent combination and coordination of multiple sensor types. Through several case studies, we illustrate the capabilities of this approach for enabling marine robots to become more reliable and flexible, improving their ability to safely navigate, detect objects, and map environments. While the implementation and experimental validation of this framework are reserved for future work, we argue that BTs can significantly boost the autonomy and resilience of underwater robots, leading to smarter and more capable systems for exploring and working underwater.