The latest fatality involving a self-driving Uber auto highlights the reality that the technology is still not all set for ubiquitous adoption. One reason is that there aren't many places where self-driving cars can actually drive.
The academic institution's Computer Science and Artificial Intelligence Laboratory (CSAIL) has developed the framework, which replaces 3D maps with Global Positioning System data and sensors. That means large portions of the United States, from Vermont's White Mountains to California's Mojave Desert, are not ready for self-driving cars.
So MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) developed a solution that would allow AVs to navigate their way without using 3D maps.
MapLite, as they dubbed their system, uses the GPS data to get a rough estimate of where on the road the vehicle is. So far, it's been successfully tested using a specially-outfitted Toyota Prius, on a number of unpaved country roads around Devens, Massachusetts.
The paper, which will be presented in May at the International Conference on Robotics and Automation (ICRA) in Brisbane, Australia, was co-written by Ort, Rus, and PhD graduate Liam Paull, who is now an assistant professor at the University of Montreal. Consider how you yourself get around: If you're trying to get to a specific location, you probably plug an address into your phone and then consult it occasionally along the way, like when you approach intersections or highway exits.
On the other hand, roads that are unpaved or not very well-lit can be extremely hard to map, so it could be a long time before 3D maps are designed for them. Prevalent systems are still more reliant on maps and only use vision algorithms and sensors to avoid dynamic objects such as other cars and pedestrians. The new system uses the combined technology of Google Maps GPS data for navigation along with LIDAR and IMU sensors for determining distances. It sets itself a final destination as well as a "local navigation goal" within sight of the car's sensors. The GPS gives the system a basic idea of the vehicle's location, while the sensors let it "see" the area around the auto, determining likely road edges based on the assumption that the road is flatter than the surrounding area. MapLite can do this without physical road markings by making basic assumptions about how the road will be relatively more flat than the surrounding areas. Currently, so many self-driving cars are tested in cities because there is an abundance of detailed 3D mapping data that not only lets the cars know where all the roads go, but how high things like curbs are.
The researchers developed a system of models that are "parameterized", meaning they describe multiple situations that are fairly similar.
Driving licences and passports compromised in 2017 security breach, Equifax now admits
Also mentioned were the TaxID data and Driver's license state data with 97,500 people and 27,000 people losing their information respectively.
MapLite is different from other map-less driving strategies that are more dependent on machine learning in that it is trained on data from one set of roads and is then tested on the others.
"At the end of the day we want to be able to ask the auto questions like 'how many roads are merging at this intersection?'" said Ort.
Yet, MapLite has some limiting factors.
MapLite is still limited and can not be relied to drive on mountain roads as it is not yet prepared to understand the dramatic changes in elevation.
Next, the researchers will expand upon the types of roads MapLite can navigate. Ultimately they aspire to have their system reach comparable levels of performance and reliability as mapped systems but with a much wider range. LIDAR is used to detect the road's edges. So rural areas that have fewer of these kinds of roads could be among the last to benefit from AVs, even if they might be among the communities that need them the most.
And that's a very big step in the right direction if you want to see self-driving cars everywhere.