You Can Help Train NASA Mars Rovers to Avoid Danger From Your Sofa

You can now participate in a project that will help NASA's Mars rovers better navigate the Martian surface from the comfort of your own home.

As part of the project, called AI4Mars, individuals who are interested in contributing can help teach an artificial intelligence algorithm to recognize geological features in images snapped by Perseverance and other Mars rovers.

Mars' terrain can be hazardous for rovers. NASA's Opportunity, Curiosity and Spirit rovers have all experienced getting stuck in sand at some point.

The goal of the AI4Mars project is to help NASA rovers identify dangerous terrain by themselves, using machine learning technology that is used in self-driving cars on Earth.

In order to achieve this, the algorithm needs lots of data to learn from. And this is where citizen scientists can play an important role.

The project asks members of the public to label features in images captured by NASA's Mars rovers, which will lead to the creation of the first open-source navigation dataset for the Red Planet.

This dataset will help improve the ability of rovers to identify different types of terrain, and determine whether or not that are in danger.

If you would like to help out, please follow this link. Volunteers will be asked to draw polygons in Mars rover images to label different terrain types, including sand, soil, bedrock and big rocks.

AI4Mars launched last year, using images taken by NASA's Curiosity Rover, which is currently exploring Mars at the same time as Perseverance.

Members of the public labeled nearly half a million images with the help of a tool that NASA's rover operators use to highlight features like sand and rock when planning driving routes on the Martian surface.

The result of this effort was an algorithm called SPOC (Soil Property and Object Classification) that was capable of correctly identifying features like this in almost 98 percent of cases.

Then NASA announced on Tuesday that images taken by Perseverance can now be used to train the algorithm.

The algorithm is still in development, but the AI4Mars team hope that it can be used to help future Mars rovers be smarter and safer, and perform even more autonomous driving than is currently possible with Perseverance's AutoNav technology.

The data is also helping the team to develop a terrain classifier that could help the current Perseverance and Curiosity rover mission. While this terrain classifier will not be on board these rover, which are currently exploring the martian surface, it could help NASA operators to plan safer routes for them.

Machine-learning algorithms essentially become smarter over time as they receive more and more data.

"Machine learning is very different from normal software," Hiro Ono, an AI researcher from NASA's Jet Propulsion Laboratory who led the development of AI4Mars, said in a statement.

"This isn't like making something from scratch. Think of it as starting with a new brain. More of the effort here is getting a good dataset to teach that brain and massaging the data so it will be better learned."

NASA’s Perseverance rover
The robotic arm of NASA’s Perseverance rover is visible in this image used by the AI4Mars project. Participants label different features to help train an artificial intelligence algorithm that will help the navigation capabilities of Mars rovers. NASA/JPL-Caltech