Let's Arm Wrestle for Sniffing that Rock

Went over the PARC yesterday to hear Kanna Rajan of the Computational Sciences Division of the NASA Ames Research Center discuss "From Interplanetary Cruise to the Surface of Mars - The Challenges of infusing AI in Space".

Integrating AI into anything has always been a tough proposition, because the generalized systems solutions are always in search of a problem that can't be solved by breaking down the problem into simpler components or by the use of sheer massive computational grit. So it was interesting to hear it used - not for complicated analysis of navigation in spaceflight, for example, as intended but instead to resolve scheduling disputes between teams sharing time on an interplanetary robot by using a "mission-critical AI application on a NASA science mission".

Through the use of a "temporal constraint network" (in other words, time-based schedule online that won't let you do conflicting things), the "Mixed-initiative Activity Plan GENerator (MAPGEN)" found a home by gluing on the standard interface used by ops at JPL for mission planning.

What does it do. In Mr. Rajan's words, it "provides a ground-based decision-support system in the critical part of the uplink command cycle for the 2003 Mars Exploration Rovers." With a limited set of commands preloaded and sent to the Rover and no operations at night allowed (power-constraints), schedules can't be reworked. Also, if there were a set of commands that turned out to be contradictory in nature, time could be lost as the Rover spins its wheels (metaphorically or actually).

So mission ops use this AI mechanism to successfully integrate all the conflicting science groups and agendas with the physical realities of working a robot at a planetary distance.