No More Blind Spots: Learning Vision-Based Omnidirectional Bipedal Locomotion for Challenging Terrain

TL;DR

This paper introduces a novel vision-based learning framework for omnidirectional bipedal locomotion that operates effectively on challenging terrains, combining simulation and real-world validation with reduced computational costs.

Contribution

It presents a new training approach that combines a blind controller with a teacher-student policy, using data augmentation to accelerate training and enable robust, omnidirectional walking.

Findings

Successful simulation and real-world tests of omnidirectional locomotion

Training acceleration by up to 10 times with data augmentation

First demonstration of vision-based omnidirectional bipedal walking

Abstract

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.