A Modal Operations Paradigm for Robust Vision-based Astronaut Following

摘要

Scenarios for manned missions to the Moon or Mars call for astronaut extravehicular teams to be accompanied by autonomous or semi-autonomous rovers. These rovers must be able to safely follow the astronauts over a variety of terrain with minimal assistance. We use a commercial color-based vision system as a high-speed sensory approach for tracking the astronaut and present a quantitative analysis of its accuracy based on a spectrum of known motion profiles. Visual tracking data is transformed into a sequence of waypoints that allow a rover to accurately follow the safe path the astronaut has previously traversed. To promote robust operation during exceptional situations, a modal operations paradigm is presented that enables the rover to recover autonomously when possible and with minimal supervision when required. We have outfitted a mobile robot with a camera connected to our color tracking system and characterized its ability to follow the astronaut target over a variety of motion profiles under nominal and exceptional circumstances. Results show the rover is able to autonomously follow and re-acquire the astronaut in most cases and is able to safely stop and request assistance otherwise.