Session
Weekend Poster Session 1
Location
Utah State University, Logan, UT
Abstract
Complex distributed systems are critical to scientific discovery in space. Future missions will take place in increasingly remote planetary environments where human intervention will neither by feasible, nore scalable toward fleet-wide mission control. To this end, autonomous onboard fault mitigation will be necessary. The unique topology of fleet systems offers opportunities for high-level contextual understanding of faults and coordinated fault mitigation not possible for single agents. We present a framework that augments single-agent fault mitigation with the context provided by a fleet. Multi-Agent Anomaly Detection (MAAD) operates on time-series sensor data to build a unidimensional distribution against which we can compare individual agents in order to detect faulty sensing hardware.
A Framework for Multi-Agent Fault Reasoning in Swarm Satellite Systems
Utah State University, Logan, UT
Complex distributed systems are critical to scientific discovery in space. Future missions will take place in increasingly remote planetary environments where human intervention will neither by feasible, nore scalable toward fleet-wide mission control. To this end, autonomous onboard fault mitigation will be necessary. The unique topology of fleet systems offers opportunities for high-level contextual understanding of faults and coordinated fault mitigation not possible for single agents. We present a framework that augments single-agent fault mitigation with the context provided by a fleet. Multi-Agent Anomaly Detection (MAAD) operates on time-series sensor data to build a unidimensional distribution against which we can compare individual agents in order to detect faulty sensing hardware.
