DribbleBot: Dynamic Legged Manipulation in the Wild
This work addresses the challenge of dynamic legged manipulation in robotics, which is incremental as it applies existing methods to a new task.
The researchers tackled the problem of enabling a legged robot to dribble a soccer ball in real-world conditions by training policies in simulation with reinforcement learning and transferring them to reality, overcoming challenges like variable ball dynamics and perception with onboard cameras, and demonstrated that current quadruped platforms are suitable for dynamic whole-body control from sensory observations.
DribbleBot (Dexterous Ball Manipulation with a Legged Robot) is a legged robotic system that can dribble a soccer ball under the same real-world conditions as humans (i.e., in-the-wild). We adopt the paradigm of training policies in simulation using reinforcement learning and transferring them into the real world. We overcome critical challenges of accounting for variable ball motion dynamics on different terrains and perceiving the ball using body-mounted cameras under the constraints of onboard computing. Our results provide evidence that current quadruped platforms are well-suited for studying dynamic whole-body control problems involving simultaneous locomotion and manipulation directly from sensory observations.