Extending a Distributed Multi-Robot Coordination Infrastructure with Predicated Broadcast and Multicast Communication

Abstract

Multi-robot systems operating in distributed environments require coordination mechanisms that are both scalable and flexible. Existing approaches commonly rely on centralized controller or undiscriminating broadcast communication, both of which introduces limitations: centralization creates single points of failure and communication bottlenecks, while broadcast-based coordination results in unnecessary message processing by robots for whom a given task is irrelevant. This thesis addresses these limitations by extending an existing ROS2-based decentralized coordination infrastructure developed at Chalmers with two new predicate-based communication mechanism: predicate-based broadcast and predicate-based multicast. Predicate-based broadcast enables system-wide message distribution in which only robots whose locally stored traits satisfy a specified boolean predicate respond to and act upon a coordination request. Predicate-based multicast extends this further by facilitating communication within dynamically formed subgroups of eligible robots, selected throughout a decentralized election process. Formal semantics are provided for both these mechanisms. The implementation preserves the core design principles of the inherited infrastructure, that is, decentralized decision making, agent anonymity, and global message ordering guarantees. The mechanisms was evaluated through two simulation scenarios using RViz. The results demonstrate that selective coordination can be achieved without the need for a centralized knowledge of robot capabilities or identities. It also shows that temporary multicast groups can be formed and dissolved at runtime, and that message ordering is maintained during the lifecycle of the coordination group. The results indicate that predicatebased communication is a viable approach for achieving selective coordination in decentralized multi robot systems.