No More Blind Spots: Learning Vision-Based Omnidirectional Bipedal Locomotion for Challenging Terrain
This addresses the need for agile bipedal robots in dynamic environments like cluttered indoor spaces, presenting a novel framework for vision-based omnidirectional locomotion.
The paper tackles the problem of enabling bipedal robots to navigate challenging terrain using vision-based omnidirectional locomotion, achieving effective movement with minimal reliance on expensive rendering and accelerating training by up to 10 times.
Effective bipedal locomotion in dynamic environments, such as cluttered indoor spaces or uneven terrain, requires agile and adaptive movement in all directions. This necessitates omnidirectional terrain sensing and a controller capable of processing such input. We present a learning framework for vision-based omnidirectional bipedal locomotion, enabling seamless movement using depth images. A key challenge is the high computational cost of rendering omnidirectional depth images in simulation, making traditional sim-to-real reinforcement learning (RL) impractical. Our method combines a robust blind controller with a teacher policy that supervises a vision-based student policy, trained on noise-augmented terrain data to avoid rendering costs during RL and ensure robustness. We also introduce a data augmentation technique for supervised student training, accelerating training by up to 10 times compared to conventional methods. Our framework is validated through simulation and real-world tests, demonstrating effective omnidirectional locomotion with minimal reliance on expensive rendering. This is, to the best of our knowledge, the first demonstration of vision-based omnidirectional bipedal locomotion, showcasing its adaptability to diverse terrains.