Working with the German Aerospace Center's Institute of Robotics and Mechatronics, Part-Time Scientists has added an autonomous rover navigation system with the capacity to process multiple images per second. The institute is known for a number of research projects, including the "Justin" mobile robotic system, designed to perform long-range autonomous operations—albeit on terra firma—and the Robotics Component Verification on the ISS (ROKVIS) project on the International Space Station to study robots used in space.
Asimov's navigation system uses a stereo camera to, in the absence of the Global Positioning System that can only be used on Earth, calculate in real-time its own motion, generate a 2.5-dimensional model of its surroundings, evaluate this model and pursue the path least likely to end in a collision. "A 2.5-D environment consists of flat polygons located in discrete distance bands," Faler explains. "Consider driving along a road with trees nearby and mountains in the distance. In 2.5-D, all trees and mountains are at the same distance, in one flat plane." If you were to get closer to one particular tree, for example, that tree would take on additional dimensions whereas the other trees would continue to look flat. For navigation, this amount of data is sufficient and drastically reduces the required computing power, Faler notes.
GPUs guide the way
Part-Time Scientists has several GPUs from NVIDIA Corp. at its disposal. GPU-based computers will play several key roles, including image filtering in transit through space and on the lunar surface; rapid mapping based on sensor and video data; calculating the best trajectory to the landing site; and optimizing communication with the rover despite a moon–Earth signal delay of a few seconds each way. GPU processors will also analyze radio-frequency signals on the moon, whose surface is made up of mostly metallic materials that interfere with these signals. This will help the team determine the best frequency to use for post-landing communications.
Although the details have not all been worked out, Böhme expects to integrate a GPU into the lander's computer systems and says another GPU could be installed in the rover itself. Any GPU used in Asimov would not be switched on until after the rover reaches the 500-meter milestone and begins autonomous operation. "There is a high interest in making a GPU on the rover possible, but there is a chance that due to technical reasons like weight, shielding and temperature management we might have to just stick to a GPU only on the lander," Böhme says.
Humans have not operated technology on the lunar surface in four decades, and today's electronics are much more sophisticated—and delicate—than those used in the Apollo missions, Böhme says. Like any other piece of equipment, a GPU sent up in the lunar lander or Asimov would need to be hardened, which includes shielding against radiation as well as thermal management components to protect it from the extreme lunar temperature range.
Böhme acknowledges that it would be possible to address any of the mission's needs without the use of a GPU, but emphasizes that one of his team's goals is to push the limits of available technology and even help drive down the cost of space travel. "Nothing is to be gained by a one-shot, very expensive mission to the moon if it does not act as a proxy for a whole new set of technologies," he says.