For the most part, self-driving car tests have focused on cities. Boston, Pittsburgh, Las Vegas and dozens of other cities have begun or plan on beginning self-driving pilot programs. There are reasons that companies look at cities. Cities generally have clear signage and well-maintained roads, two things that make the huge challenge of AI driving a little easier to break down and tackle.
Companies like Google or Uber also invest in complex 3D-mapping for these urban areas, labeling each and every road sign and lane. Given the complexity of the operation and the fact that millions of miles of U.S. roadsare unpaved, unlit, or unreliably marked, it’s no wonder the development focus so far has been decidedly urban.
That’s something MIT’s Computer Science & Artificial Intelligence Lab (CSAIL) wants to change.
Rather than relying on 3D mapping, MapLite focuses its efforts on satellite mapping like the kind you can find on Google Maps, and a series of LIDAR and IMU sensors.
The system then creates two goals: a final destination and a “local navigation goal,” in view of the car. The LIDAR then estimates where the road begins and ends. Without sign markings, the LIDAR makes basic assumptions that the road will be relatively more flat than the surrounding areas that are not road.
“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. A system like this that can navigate just with on-board sensors shows the potential of self-driving cars being able to actually handle roads beyond the small number that tech companies have mapped,” says CSAIL graduate student Teddy Ort, who was a lead author on a paper about MapLite, in a press statement.
The system was tested around Devens, Massachusettes, a census-designated area in central Massachusettes with a population under 2,000. Driving a Toyota Prius around on multiple unpaved country roads, the CSAIL team was successfully able to self-drive and detect the road more than 100 feet in advance.
The system isn’t perfect yet. It’s best equipped for forest-type environments like those around Devens, it can’t handle a mountain road due to their dramatic changes in elevation. But the successful test is a first step, researchers hope.
“I imagine that the self-driving cars of the future will always make some use of 3-D maps in urban areas. But when called upon to take a trip off the beaten path, these vehicles will need to be as good as humans at driving on unfamiliar roads they have never seen before. We hope our work is a step in that direction,” says Ort.