Goal-directed risk-aware autonomous explorers by Brian Charles Williams
Planetary scientists have set their appetite for an ambitious slate of missions that they’d like to see over the next decade, including a Venus lander, a tour of the Trojan asteroids, the Mars Sample Return, and even an interstellar mission.
Looking further into the future, the “holy grail” of space missions is an explorer that looks for evidence of life deep under the ice of Jupiter’s moon Europa.
Over the last decade NASA has had great success at exploring the surface of Mars using sophisticated hardware and a relatively simple operations paradigm. The rovers carry out a linear sequence of low-level tasks that are assigned to them by an operations team comprising human scientists and drivers. The rovers are protected from failure through a set of fault detection rules that largely halt the mission until operators can intervene. During each operations cycle, the operation team devotes most of its time to the tedious task of writing down these sequences, and testing that they are correct. The time required to write these sequences seriously detracts from the operator’s ability to focus on strategy and science. In addition, when sequence errors creep into rover operation, the mission is often aborted, and science is lost for the day, thus degrading science return provided by this expensive instrument. A similar paradigm is applied to most planetary and Earth orbiting missions, and back on Earth for underwater exploration, for manufacturing robots, and for some aerial drone missions. In each case, the process is time consuming and error prone. It distracts the operators from thinking strategically, can be brittle to failure, and can limit the range of the technology’s application.
Goal-directed risk-aware autonomous explorers
Published on Nov 18, 2016
Annual magazine review of MIT Aeronautics and Astronautics Department research and educational initiatives.