Stable Trajectory Planning for Quadruped Robots using Terrain Features at Feet End
Congfei Li, Shuyue Lin, Shenwei Qu, Zhuoyuan Liu, Qingjun Yang, Max Q.-H. Meng, Yuxiang Sun
AI summary
Problem
Existing planners rely on grid maps or body-centric objectives that ignore foothold-level terrain interactions, failing to exploit quadruped agility on irregular surfaces.
Approach
The method models terrain with a Sparse Variational Gaussian Process to create a differentiable elevation map, then optimizes body waypoints using CMA-ES to minimize foot-end height variations and satisfy terrain constraints.
Key results
- SVGP accurately predicts terrain features from point-cloud data
- CMA-ES optimizes body waypoints by penalizing foot-end height variations
- Simulation and real-world tests demonstrate smoother CoM paths and reduced Z-axis momentum
- Open-sourced code enables seamless integration with NMPC controllers
Why it matters
Enables quadruped robots to traverse complex, unstructured terrains more safely and efficiently, advancing agile locomotion for inspection and rescue applications.
Abstract
Quadruped robots have received increasing attention in recent years. Most existing trajectory planning algorithms for quadruped robots focus on how to avoid obstacles and achieve shortest trajectory or time, which is similar to the planning algorithms for mobile robots. These algorithms could not take full advantage of the agility and flexibility of quadruped robots. This letter designs a trajectory planner by taking advantage of the agility and flexibility of quadruped robots. With our trajectories, quadruped robots could navigate through complex terrains with more stability (e.g., less momentum variations along Z-axis). To achieve this goal, we use ground features at the landing point of the feet end to construct objective function, rather than using the center point of the robot body. Current discrete map representations, such as grid map or cost map, are difficult for optimization algorithms to introduce environment constraints. So, we use the Sparse Variational Gaussian Process (SVGP) to predict terrain features with point-cloud data as input, so that the environment constraints can be introduced into the optimization problem. Experimental results in both simulation and real-world environments demonstrate the effectiveness of our method.