Robot Doggy Paddle: Modeling and optimization reveal efficient swimming gaits

Deep water is common in many typical operating environments of legged robots. Despite this, study of legged robot gaits has mainly focused on terrestrial movement. Efficient swimming gaits are needed to greatly expand the operational capability of legged robots. Building on previous work on single leg swimming gaits, this study describes a 7 degree-of-freedom reduced-order dynamic model that captures surface swimming of a bipedal “sagittal quadruped” robot made of two rigidly connected “hip” masses with revolute-revolute legs connected at each hip. With the model, we investigated the effect of the gait parameters of stroke length and angle, leg extension and retraction, and front-rear phasing on the speed, efficiency, and stability of swimming gaits. To study the effect of body length, center-of-mass location, and buoyancy, on speed, efficiency, and stability of swimming, we used direct colocation optimization to find an optimal gait for each set of parameters. These results could be used to inform gaits and design parameters for existing and novel legged robots to extend their operational capabilities beyond terrestrial locomotion.