An integrated distributed monitoring for mission-based systems: On the example of an autonomous unmanned helicopter

This paper presents a new approach of an integrated distributed monitoring system for an autonomous unmanned helicopter. The navigation mission consists of reaching areas of interest on a path composed of specific waypoints. The task of such a monitoring system is to determine whether the helicopter behaves as intended in order to detect faults in the mission execution early and to enable a re-planning of the mission or a failure recovery in due time. This is a very important issue for autonomous systems as they have to cope not only with ordinary subsystems failures, but also with unexpected situations which make the mission goals unattainable. The approach uses a special combination of Petri Nets and Monte Carlo Methods (particle filter) and integrates different layers of the system architecture such as control actions (path-planning, low-level control), environment perception (vision-based recognition of landmarks), state observation (estimation of speed, position, attitude) and decision making (mission planning and re-planning) in a unified model.