Controlling Dynamic Gaits for Quadrupedal Robots

Bridging the gap between the skill sets of legged robotic systems and that of real animals is one of the major challenges in the field of robotics. Our long-term goal is to develop and control quadrupedal robots that are able to operate semiautonomously, i.e. they are only guided by users through highlevel velocity commands, such as move forward at 1 m/s, or turn with 0.5 rad/s. To achieve this, the robots should have a large repertoire of gaits and the ability of selecting them based on desired speed, energy efficiency considerations, or the upcoming terrain. At the Autonomous Systems Lab we recently developed StarlETH (Springy Tetrapod with Articulated Robotic Legs), an electrically driven quadruped robot [1]. StarlETH’s weight of 23kg and leg length of 0.4m correspond to the dimensions of a medium sized dog, and it uses an actuation scheme based on highly compliant series-elastic actuators that enable torque control. Our aim is to increase StarlETH’s repertoire of motions to include faster, more lifelike dynamic gaits.