Fruits of some of the practical applications of Rodney Brooks' behavior-based robotics are finding their way - albeit, in a limited fashion - into remote-controlled rovers being developed by NASA for exploring other planets and asteroids.
"[Subsumption architecture] influenced some aspects of the design of Sojourner, but Rod's system has little memory and is less knowledge-driven. We need to use more [memory] because we have a fairly detailed interactive cycle," said Jake Matijevic, manager of the Pathfinder rover at NASA's Jet Propulsion Laboratory. "Sojourner has to remember where obstacles are and to correlate specific measurements with images."
Matijevic says NASA, which funded some of Brooks' work and periodically collaborates with some of his students, based some of Sojourner's intelligence on a well-known application of subsumption architecture, hazard avoidance. In obstacle avoidance, a robotics device relies on various sensors to feed information into the system's processor, alerting it to the existence of obstacles in its path. Reactions to this information are nearly immediate, as a robot will change its path slightly, slow down, or, if an object quickly draws near, stop until the obstacle passes.
But even with Brooks' technology, NASA has been more inclined to drive Sojourner based on commands JPL scientists transmit up to it, particularly since this mission is turning up rich scientific data on aspects of the arid planet such as mineral content, explained Matijevic.
Brooks saw an otherworldly application for his work very early on. In the 1989 paper "Fast, Cheap and Out of Control: A Robot Invasion of the Solar System," Brooks and co-author Anita M. Flynn proposed different types of robots that could be used for unmanned space missions. These robots ranged from a six-legged machine that walked on its own to large rovers like Sojourner and Rocky 7 to small, autonomous micro rovers weighing 1 to 2 kilograms.
Most intriguing were the small rovers which would, Brooks and Flynn envisioned, operate as an ant colony and reduce the chances of delays and failures of missions. For example, imagine the Pathfinder mission was a journey for several micro rovers instead of just Sojourner. If one rover winds up beached upon a rock, scientists on Earth could still perform analysis and experiments with the other rovers that would be free to explore.
"The overall science component of the mission need not be compromised if time is spent rethinking the size and shape of the instruments used," Brooks and Flynn wrote in the paper which appeared in The Journal of the British Interplanetary Society.
Despite this promise, Matijevic greets these ideas with some reserve and skepticism. Calling NASA strategies for future unmanned vehicles mixed, Matijevic sees a combination of smaller, unitasking craft together with larger rovers.
"To go to asteroids and moons, we need vehicles that are limited and targeted in their function," he explained. "These will weigh more heavily on subsumption architecture."
For subsequent Mars journeys, Matijevic says NASA will look to larger rovers that will use the basic tenets of Brooks' behavior-based AI underneath a more sophisticated control scheme. This is all the better to ensure that researchers get the scientific data they're after.
