AI will soon be driving Martian rovers
Humans may not have reached Mars yet, but AI has, in the form of the targeting software currently used by the Mars Science Laboratory’s Curiosity rover. Scientists, Chien explains, use artificial intelligence to specify very high-level target requests, such as searching for rocks with a set texture, size or distribution. The software then determines the right orientation for aiming NASA’s Chemistry and Camera (ChemCam) laser at the targets and fires it, burning the planet’s ever-present iron dust off their surfaces in order to expose their underlying mineral structures.
The next generation of rover, however, will be able to do much more, programmed not only to drive itself, but to plan its own tasks. Currently, human scientists direct what the rover does and when. The 2020 version will be able to analyse data itself, allowing it to work out how much time to spend on a task. If it completes one early, the rover will either move on or decide to spend higher-quality time on the original assignment, for instance taking a 4x4 image mosaic instead of the 3x3 scientists requested.
“We’re very excited about that,” Chien says, and not just because of the obvious increase in rover efficiency. By allowing some autonomy in future missions, multiple rovers will be able to team together, exploring not only Mars’s surface, but its caves. They’ll connect through a dynamic allocation algorithm, a computer system technique for allocating memory. This will help the group of rovers to decide which part of the cave each will map. Then when one rover dies and can’t send feedback to the team, others can take over.
“The next generation of Mars rover will be able to do much more – not only driving itself, but planning its own tasks”
An artist’s impression of NASA’s Mars 2020 Rover deploying its robotic arm. AI will help it make its own decisions during tasks
Curiosity’s photograph of drilling sites on Mars – but future rovers will decide how to use their own time