Quadruped robot walks complex terrain using raw LiDAR data
PointWalker: Mapless Quadruped Locomotion Directly From Raw LiDAR Point Clouds
Robotics
Summary
Robots with four legs can have trouble walking on tricky surfaces like stairs or gaps. Many techniques rely on creating maps or images to understand where to step, but these can fail if the robot's position is off or the environment is tricky to see. The authors created a new method called PointWalker that skips making maps and instead uses raw 3D points from laser sensors directly to decide how to walk. Their approach works in many different terrains and was tested successfully both in simulations and on a real robot carrying heavy loads.
What this means in practice
- •For field robotics teams: Enable quadruped robots to traverse diverse rugged terrains without building or relying on terrain maps, improving reliability in outdoor inspection missions.
- •For robotics hardware integrators: Deploy quadruped robots that interpret raw LiDAR data directly for navigation, simplifying sensor setups and reducing dependency on odometry for balance and stepping.
Authors
Sicen Li, Zhen Chu, Chao Li, Jun Zheng, Qiuguo Zhu, Jun Wu
Abstract
Quadruped robots deployed for inspection and field missions must cross heterogeneous terrain, from stairs, slopes, and gaps to high platforms, stepping stones, and balance beams, where safe support is sparse or see-through. Existing methods interpose a terrain map or a depth image between sensing and control. Maps are costly to maintain, corrupted by odometry drift and outliers, and misled by see-through gratings, while depth cameras degrade with lighting and surface material. In this paper, we present PointWalker, a mapless framework whose single reinforcement learning policy directly consumes raw point clouds from four light detection and ranging (LiDAR) sensors. Voxelized points are encoded by linear self-attention and fused with proprioception through cross-attention, so terrain geometry reaches the policy without maps or foothold estimation. One proximal policy optimization (PPO) run, with terrain-aware rewards and dual-channel simulation of see-through gratings, yields one policy for all seven terrain classes with a 90.1% mean success rate in simulation. Deployed zero-shot on a Deep Robotics X30 at 50 Hz onboard, it crosses 0.8-m gaps, climbs 0.8-m platforms and 45° steel-grating stairs, carries a 50-kg payload, and completes a 2.5-km mountain route.