Abnormality Detection in Multiagent Systems Inspired by the Adaptive Immune System

Fault tolerance is one of the most prominent challenges in the field of multirobot systems. The efficient and long term operation of a robot collective requires an accurate detection and accommodation of abnormally behaving robots. Most of the existing fault tolerant systems prescribe a characterization of normal behavior, and train a model to recognize them. Behaviors not recognized by the model are labelled abnormal. However, these models require a priori knowledge of the normal behavior. Furthermore, multirobot systems employing these models do not transition well to scenarios involving temporal changes to normal behavior. We propose to address this challenging problem by taking inspiration from the regulation of tolerance and (auto)immunity in the adaptive immune system. We adopt the Crossregulation model, used to explain the robust immunological maintenance of tolerance, and deploy it within a multiagent system. Results of extensive simulation-based experiments demonstrate that a distributed multiagent system can detect abnormalities under varying conditions of normal behaviors. The collective dynamics gives rise to a meaningful normalabnormal classification of the behavior by individual agents, even if these categories were not prescribed a priori in the agents.