My primary research focus is to enable planetary rovers to autonomously select long-range paths considering mission objectives, available resources, operational constraints and uncertainty. The TEMPEST planner uses incremental search to derive optimal global plans that achieve the objectives while enforcing global constraints. It integrates several models of the planetary mission domain: a world model - terrain, planetary motion, solar flux, dynamic line-of-sight geometry; a rover model - locomotion, power collection and consumption, and sensor fields-of-view; an action model - driving, stationary battery charging, hibernation, science activities; a constraint model - geometric, temporal, and resource-related; and a model of the mission objectives - the start specification, and specifications of goal positions, legal time and battery energy ranges and activities to be conducted. Plans are time-synchronized, energy-optimal routes between goals.
A preliminary, offline, open-loop version of this software was demonstrated in July 2001 on the Hyperion rover in the Canadian Arctic, where it generated multi-kilometer circuits that enabled solar-powered traverses lasting 24 hours. In April 2003, Hyperion employed an online version of TEMPEST in the Atacama Desert in Chile that was able to plan and re-plan routes on the fly to single targets several kilometers apart. Most recently, in September and October of 2004, TEMPEST was run aboard the Zoe rover. In conjunction with an intelligent executive process and goal manager processes, TEMPEST successfully derived plans that integrated navigation between several scientist-designated targets, science activities at those locations, and battery energy management for multi-kilometer regional investigations. In the coming year, the system will aid Zoe in its bid for extended autonomous operations in support of a remotely-directed search and characterization of life in the Atacama desert.