Trust-Based Information Propagation on Multi-robot Teams in Noisy Low-Communication Environments

Abstract

One of the challenging requirements of large multi-robot systems is scalable communication methods that are robust to noisy low-communication environments. On a multi-hop, wireless sensor network, it may be desirable for a node to issue a system-wide alert on detection of an event of interest. However, if there is a high false-positive rate, then system-wide false alarms may frequently occur. Giant honeybees (Apis dorsata) generate large-scale spiral waves triggered by the presence of a threat. Studies show that the initial seed of the waves and the re-transmission behavior are non-trivial and thus may be specially tuned for this low-communication scenario. Motivated by these adaptive patterns in giant honeybees, we develop a distributed approach for sharing awareness of critical events that is able to damp the propagation of false-alarm signals. We validate the algorithm’s performance for a WSN detecting a hostile UAV in the SCRIMMAGE simulation framework.