Pass the Parcel: A Relay-Based Nonprehensile Approach to Manipulate Cargo With Multirobot Systems

This article presents a relay-based strategy for multirobot object transportation in constrained environments, where an object is passed sequentially between agents. To enable physical handovers using simple differential drive platforms, we employ nonprehensile pushing, which introduces strict kinematic and friction constraints. We propose an integrated, manipulation-aware framework that jointly optimizes the locations of relay points and their connecting pushing trajectories. Our assignment method explicitly accounts for handover pose constraints and exploits push-face switching to minimize turning effort. The connecting trajectories are generated via a decoupled approach: geometric paths are synthesized using curvature bounded primitives within safe corridors to guarantee pushing stability, followed by temporal optimization to balance robot workloads. To bridge offline planning with online execution, each robot runs a decentralized behavior tree featuring short horizon replanning and spatial–temporal coordination for smooth, collision free handovers. Extensive real world experiments and quantitative comparisons against baselines validate the effectiveness, reliability, and practical viability of the proposed approach.