Overview
This study presents a decentralized multi-robot system, designated R2P2 (Roles with Rules and Proportional-control Primitive), designed for the collaborative transportation of rectangular boxes. The approach focuses on managing the challenges associated with moving objects over diverse terrains, specifically those with varying inclination and friction properties. R2P2 operates asynchronously, aiming to reduce communication and synchronization requirements while mitigating single points of failure in multi-robot systems.
Research Context
Collaborative object transport by pushing using multiple robots holds relevance in various applications, including construction, warehouse logistics, and post-disaster debris clearing. A key challenge in these scenarios is achieving effective collaboration across surfaces possessing different inclination and friction characteristics.
Approach
R2P2 is an asynchronous and decentralized task and motion planning framework. It is designed for transporting rectangular boxes of varying mass over flat, uphill, and downhill terrain. The core mechanism involves assigning specific roles to individual robots based on a set of rules. These roles are determined by the mode of manipulation required, classifying actions as either related to box rotation or translation.
Following role assignment, the control of robot velocity is implemented through either rule-based control or proportional control, depending on the assigned role. Each robot within the R2P2 framework is assumed to have the capability to observe its own location and heading, as well as the location and heading of the box being transported, to execute its assigned role and associated controls.
Findings
The R2P2 approach was evaluated using a six-robot team within a simulator constructed using NVIDIA IsaacSim. This evaluation demonstrated the generalizability of R2P2 across different scenarios involving varying surface friction, surface inclination, and box mass. The simulation results indicated that R2P2 achieved a better success rate when compared to a standard virtual-leader-follower method.
Further validation of R2P2 included a physical experiment. In this experiment, the system was implemented onboard four TurtleBots. These robots were tasked with moving a 1.2 kg box, and the experiment successfully demonstrated the execution of the R2P2 approach in a physical setting.
Potential Applications
The collaborative transport capabilities demonstrated by R2P2 could be applied in various real-world contexts. These include areas such as construction, where multi-robot systems might assist in moving materials; warehouse environments for automated item relocation; and post-disaster situations, potentially aiding in debris removal operations.
Key Limitations Mentioned by Researchers
The source document does not explicitly discuss key limitations of the R2P2 system as mentioned by the researchers.