Learning Active Spine Behaviors for Dynamic and Efficient Locomotion in Quadruped Robots
This addresses the challenge of dynamic and efficient locomotion for quadruped robots with active spines, representing an incremental improvement in robotic control.
The paper tackled the problem of achieving fast quadrupedal locomotion with an active spine, which is complex and has seen limited success, by using a deep-reinforcement learning framework in simulation, resulting in a bounding speed of 2.1 m/s with a maximum Froude number of 2, improved stride length, cost of transport, and realistic natural frequency.
In this work, we provide a simulation framework to perform systematic studies on the effects of spinal joint compliance and actuation on bounding performance of a 16-DOF quadruped spined robot Stoch 2. Fast quadrupedal locomotion with active spine is an extremely hard problem, and involves a complex coordination between the various degrees of freedom. Therefore, past attempts at addressing this problem have not seen much success. Deep-Reinforcement Learning seems to be a promising approach, after its recent success in a variety of robot platforms, and the goal of this paper is to use this approach to realize the aforementioned behaviors. With this learning framework, the robot reached a bounding speed of 2.1 m/s with a maximum Froude number of 2. Simulation results also show that use of active spine, indeed, increased the stride length, improved the cost of transport, and also reduced the natural frequency to more realistic values.