Distributed Service Discovery over Heterogeneous Robotic Systems-of-Systems

In robotic swarms, nodes must cooperate in order to accomplish complex tasks. Robot interactions may occur at run-time based on the dynamic network topology and on mission-specific goals. For this reason, service discovery mechanisms are fundamental in order to inform each robot about the functionalities offered by the other peers. Although several network discovery protocols have been proposed so far, none of them fits the unique characteristics of robotic swarms in terms of dynamic topology, communication heterogeneity, and latency requirements. In this paper, we fill such gap by designing, implementing and evaluating a novel service discovery mechanisms for generic Robotic Systems-of-Systems (RSoS). The study proposes algorithmic and practical contributions. Regarding the first, we describe a novel distributed service discovery algorithm for RSoS which adapts to highly mobile robotic environments while limiting the network overhead and latency. Regarding the latter, we implemented the proposed algorithm within the Uhura swarm robotics framework; as a result, our solution is able to support multi-radio scenarios where robots are provided with heterogeneous Machine-to-Machine (M2M) communication technologies. In addition, we validate our solution through large-scale simulations and a test-bed in which ground robots are able to discover a Federated Learning (FL) task and join it at run-time to improve the accuracy.