Distributed Event-Triggered Localization for High Latency Communication

This paper presents a Kalman Filter-based solution to the problem of decentralized cooperative robot localization and tracking with high latency and low bandwidth communication using an event-triggered fusion algorithm. Event-triggering is a compression technique for measurement sharing whereby only ‘surprising’ measurements are exchanged between agents. For unsurprising measurements, the lack of a measurement transmission is itself information about the value of a measurement that can be fused by other agents. This relies on consistency between ‘common’ estimates of agents, but, in high latency communication, consistency is not guaranteed and is difficult to achieve with bandwidth restrictions. We present a novel ‘Delta-Tier’ algorithm to address these issues, using token-passing to preserve consistency and auto-selection of event trigger thresholds δ to fit the available bandwidth. We demonstrate the success of our approach in achieving decentralized performance that is on par with idealized centralized fusion for an autonomous underwater robotics application in a simulated nonlinear environment, where robots must localize each other and track other non-cooperative agents.