Motion Planning for Multimodal Underwater Legged Robots

This paper presents a method for enabling autonomous navigation of a legged underwater robot designed to perform targeted exploration missions on the seafloor. Unlike traditional underwater vehicles, which rely on thrusters or tracks and often struggle with stability and precise movement on uneven terrain, the proposed system leverages a hybrid locomotion strategy-including walking, jumping, swimming, and sinking gaits-to adapt to diverse and unstructured environments while minimizing disturbance to the seafloor. Experimental results demonstrate that the robot is capable of generating multi-modal trajectories, dynamically choosing between swimming, jumping, descending and walking gaits. Both local and global motion plans are generated to achieve either distance or energy optimal motions around or over seafloor obstacles.