MIT system enables autonomous driving on poorly mapped roads


A new system from MIT, known as MapLite, enables self-driving cars to navigate using just GPS and sensors.

Self-driving systems are typically seen being tested in urban environments – locations that have been mapped and manually labelled to allow an AI to navigate the roads safely.

However, a team of researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), supported by the Toyota Research Institute, have rigged a Toyota Prius with LIDAR and IMU sensors and used CSAIL’s MapLite system to enable the vehicle to drive autonomously beyond city limits.

Rural roads often lack the painted lines, signs, and curbs that are used to map and autonomously navigate in urban areas, which is why huge stretches of the US are currently off-limits to self-driving vehicles, at least in their autonomous settings.

“The cars use these maps to know where they are and what to do in the presence of new obstacles, like pedestrians and other cars,” says Daniela Rus, director of CSAIL. “The need for dense 3D maps limits the places where self-driving cars can operate.”

MapLite autonomous driving
MapLite gets around this by doing away with the need for 3D maps, instead combining the sparse topological maps used for GPS navigation (for global navigation) with sensor-base perception (for local navigation).

The MapLite research paper, ‘Autonomous Vehicle Navigation in Rural Environments without Detailed Prior Maps’, describes how the system works:

First, a local navigation goal within the sensor view of the vehicle is chosen as a waypoint leading towards the global goal. Next, the local perception system generates a feasible trajectory in the vehicle frame to reach the waypoint while abiding by the rules of the road for the segment being traversed.

LIDAR is used to create a 3D point cloud, allowing the system to approximate the edges of the road, as well as a more reliable filtered estimate, meaning the road can be accurately detected over 100 feet ahead.

“The reason this kind of ‘map-less’ approach hasn’t really been done before is because it is generally much harder to reach the same accuracy and reliability as with detailed maps,” said CSAIL graduate student Teddy Ort.

“A system like this that can navigate just with onboard sensors shows the potential of self-driving cars being able to actually handle roads beyond the small number that tech companies have mapped.”


Related posts

Leave a Comment