Learning to Behave: Adaptive Behavior for Planetary Surface Rovers
Stefan Schaal, Auke Jan Ijspeert, Aude Gemma Billard, Sethu Vijayakumar · 2004
Robotic missions to planetary surfaces are becoming more ambitious and of longer duration. The nominal mission timeline for the MER (Mars Exploration Rovers) called Spirit and Opportunity currently on the Martian surface is 90 days, with extensions to 180 days and beyond depending on rover health. The upcoming 2009 MSL (Mars Science Laboratory) mission is planned to be 300- 500 days, and will possibly involve traverses on the order of multiple kilometers. Due to time delays of up to 40 minutes round-trip for control, the rovers will require a high degree of onboard autonomous behavior that must also adapt to declining health and unknown environmental conditions during a long duration mission. This paper presents an algorithm for onboard adaptive learning of weights within a rover hierarchical behavior control framework called SMART (System for Mobility and Access to Rough Terrain). SMART is based on earlier work in free flow behavior hierarchies for planetary surface rovers (Huntsberger & Rose, 1998; Huntsberger, 2001). We also present the results of some preliminary laboratory and field studies.