Last week, Land Rover announced a new project called “Cortex,” which will cost about $5 million. The goal of this project is “to create self-propelled robot vehicles capable of independently navigating off-road in all weather conditions. Of course, the first variants of Robo Rover will not be able to overcome water obstacles and perform other extreme driving tricks on their own in the near future, those off-road robots that will be the end result of Cortex programs will be able to do so.
The problems in the implementation of automatic driving in off-road conditions are that there are no clear-cut pegs, which in normal conditions are automobile roads. Concepts such as lanes, boundary lines and, in some cases, no distinction between the road and what is next to it are completely absent. In such conditions, a robot car simply has nothing to “lock onto” while driving, but now a group of 20 specialists from the University of Birmingham and Myrtle AI are working to solve the tethering problem by developing a new specialized artificial intelligence system.
For now, the researchers are focusing their efforts on a radar system that is already partially used in cruise control systems in production cars. In production cars, the system processes no more than 10 percent of the data transmitted to it by radar, highlighting only the data that relates to moving objects in the immediate vicinity. But a future robot SUV will need to process at least 90 percent of the incoming data to navigate the terrain, and the robot car’s radar must have even greater resolution. And the vehicle’s control system, in turn, must be able to process gigabytes of data per second.
“First and foremost in the Cortex program, we will be developing new algorithms and clever data processing methods that can use as much of this data as possible,” says Nigel Clark, project manager, “The core of the new computing system will definitely be based on deep learning and self-learning principles.”
The first phase of the Cortex project will last 30 months, and no one expects robot SUVs to have time to cover millions of kilometers in a variety of conditions during that time. Land Rover plans to use a systematic approach and move towards the target, as they say, slowly but surely.
