Collaborative Quadruped Transportation in 3D Terrain with Constrained Diffusion
Williard Joshua Jose, Li Chen, Hao Zhang
AI summary
Problem
Existing multi-robot transportation methods struggle to reconcile team-level trajectory planning with individual robot control while accounting for payload-induced kinematic constraints, anisotropic velocity limits, and unstructured 3D terrain obstacles.
Approach
CQTD formulates the task as a constrained bilevel optimization problem where an upper-level diffusion model generates terrain-aware team trajectories, and a lower-level optimizer computes individual velocity commands that satisfy closed-chain kinematics, collision avoidance, and robot-specific velocity limits.
Key results
- Enables closely-coupled dual-quadruped payload transport across unstructured 3D terrain
- Introduces a constrained bilevel optimization framework integrating diffusion-based team planning with individual velocity control
- Develops and releases an open-source Gazebo simulation with automatic 3D terrain generation and ROS-based dual-quadruped control
- Demonstrates superior performance over baselines in high-fidelity simulations and real-world quadruped robot teams
Why it matters
Provides a scalable, constraint-aware framework for coordinated multi-robot payload transport in unstructured outdoor environments critical for search-and-rescue and logistics.
Abstract
Recently, multi-robot systems have gained sig- nificant attention for their promise of scalable efficiency, reliability, and cost savings. A crucial capability is collaborative transportation, where a team of robots works together to transport a payload. However, key challenges remain, such as potential conflicts between team-level decisions and individual- level robot controls, team kinematic constraints imposed by the robot-payload coupling, and diverse obstacles encountered in 3D terrain. We present Collaborative Quadruped Transportation with Constrained Diffusion (CQTD), enabling a team of closely coupled quadruped robots to collaboratively transport a payload across 3D terrain. A diffusion-based upper level learns terrain-aware team-level trajectories satisfying team kinematic constraints due to the payload coupling, while a lower level optimizes velocity controls of individual robots satisfying collision and anisotropic velocity constraints. Experiments in high-fidelity simulations and on real-world quadruped robot teams demonstrate that CQTD outperforms baseline methods in challenging 3D terrain scenarios requiring closely-coupled collaboration between the quadruped robots. More details of this work are available on the project website: https://hcrlab.gitlab.io/project/cqtd.