BY ADAM MANN
A new system could make space exploration robots faster and more efficient by predicting where they will be in the very near future.
The engineers behind the program hope to overcome a particular snarl affecting our probes out in the solar system: that pesky delay caused by the speed of light. Any commands sent to a robot on a distant body take a certain amount of time to travel and won’t be executed for a while. By building a model of the terrain surrounding a rover and providing an interface that lets operators forecast the how the probe will move around within it, engineer can identify potential obstacles and make decisions nearer to real-time.
As an example, the distance between Earth and Mars creates round-trip lags of up to 40 minutes. Nowadays, engineers send a long string of commands once a day to robots like NASA’s Curiosity rover. These get executed but then the rover has to stop and wait until the next instructions are beamed down.
Because space exploration robots are multi-million or even multi-billion-dollar machines, they have to work very carefully. One day’s commands might tell Curiosity to drive up to a rock. It will then check that it has gotten close enough. Then, the following day, if will be instructed to place its arm on that rock. Later on, it might be directed to drill into or probe this rock with its instruments. While safe, this method is very inefficient.
“When we only send commands once a day, we’re not dealing with 10- or 20-minute delays. We’re dealing with a 24-hour round trip,” said Norris.
Norris’ lab wants to make the speed and productivity of distant probes better. Their interface simulates more or less where a robot would be given a particular time delay. This is represented by a small ghostly machine — called the “committed state” — moving just ahead of a rover. The ghosted robot is the software’s best guess of where the probe would end up if operators hit the emergency stop button right then.
By looking slightly into the future, the interface allows a rover driver to update decisions and commands at a much faster rate than is currently possible. Say a robot on Mars is commanded to drive forward 100 meters. But halfway there, its sensors notice an interesting rock that scientists want to investigate. Rather than waiting for the rover to finish its drive and then commanding it to go back, this new interface would give operators the ability to write and rewrite their directions on the fly.
The simulation can’t know every detail around a probe and so provides a small predictive envelope as to where the robot might be. Different terrains have different uncertainties.
“If you’re on loose sand, that might be different than hard rock,” said software engineer Alexander Menzies, who works on the interface.
Menzies added that when they tested the interface, users had an “almost game-like experience” trying to optimize commands for a robot. He designed an actual video game where participants were given points for commanding a time-delayed robot through a slalom-like terrain. (Norris lamented that he had the highest score on that game until the last day of testing, when Menzies beat him.)
The team thinks that aspects of this new interface could start to be used in the near future, perhaps even with the current Mars rovers Curiosity and Opportunity. At this point, though, Mars operations are limited by bandwidth. Because there are only a few communicating satellites in orbit on the Red Planet, commands can only be sent a few times a day, reducing a lot of the efficiency that would be gained from this new system. But operations on the moon or a potential asteroid capture and exploration mission – such as the one NASA is currently planning – would likely be in more constant communication with Earth, providing even faster and more efficient operations that could take advantage of this new time-delay-reducing system.