R2P2 algorithm lets robot teams transport boxes over varied terrain

Researchers from the University at Buffalo have introduced R2P2 (Roles with Rules and Proportional-control Primitive), a decentralized task and motion planning algorithm for collaborative box transport by multiple robots. Unlike centralized approaches, R2P2 assigns each robot a role—push, support, or prevent—based on real-time rules that consider the terrain (flat, uphill, or downhill) and the required manipulation mode (rotation or translation). Robots then execute role-specific velocity control using either rule-based or proportional control, requiring only local observations of the box and self-state. This design eliminates communication bottlenecks, synchronization overhead, and single points of failure, making the system highly robust and scalable.

In simulation using NVIDIA IsaacSim, a six-robot team demonstrated the algorithm's generalizability across different surface friction coefficients and inclinations, consistently achieving higher success rates than a standard virtual-leader-follower method. The approach was also validated physically with four TurtleBot robots transporting a 1.2 kg cardboard box across various surfaces. The results confirm that R2P2 can handle real-world uncertainties while maintaining efficient coordination. Potential applications include warehouse logistics, construction site material handling, and post-disaster debris cleanup—scenarios where terrain variability and robot failures are common challenges.