Living in the country has a lot to recommend it — peace and quiet, beautiful scenery, access to hunting and fishing and the space to farm are just a few of them. But it has its drawbacks, too, and among them is the lack of public transport, taxis and ride sharing services. While ride share companies are rolling out self-driving buses and ride share cars in metro areas around the country, these vehicles don’t have the complex 3D maps they need to function in rural areas.
However, that could soon change. MIT researchers are working on new mapping tech that could solve many navigational problems with self-driving cars, including helping them cope with road work, poor weather conditions and country roads. And that means that rural Americans won’t be left out of the ride share revolution — and it also means that residents of more populous areas will soon be able to rely on self-driving ride shares that know how to cope with sudden obstacles and changes in their pre-existing maps.
Why Self-driving Cars and Country Roads Don’t Mix
Why can’t self-driving cars cope with road work or rural roads or the sudden appearance of a pedestrian in their path? It has to do with how they navigate. Most self-driving cars in urban areas rely on a 3D map to get around urban streets. That’s all well and good — until the route ahead suddenly doesn’t match the map, as it might when road conditions change due to ice or snow or when construction happens or when potholes appear.
It’s relatively easy for self-driving car manufacturers to build these maps for well-lighted, well-marked urban roads. That’s why the first driverless ride sharing services have rolled out in major cities. Driverless buses have already hit the road in several cities, including Las Vegas, and driverless ride sharing with companies like Uber is available in some metro areas. But the 3D maps these vehicles use don’t update in real time; they have to be painstakingly created before the fact, and that means they may not reflect road work or changes in road conditions.
It’s also more difficult and less cost-effective for manufacturers to make such maps for rural areas, where roads are dark and often poorly marked or unmarked. Road conditions can change much more rapidly in rural areas, where snow is removed and damage is repaired less frequently.
Bringing Driverless Cars to the Countryside
The solution? A driverless car that can see the road ahead, estimate where the sides of the road or lane are and create its own map of the route as it goes along. That’s exactly what MIT researchers Paul Ort, Liam Paull and Daniela Rus have created. Their new navigational system for self-driving cars uses LIDAR, a type of radio sensor that uses lasers instead of radio waves, to create detailed 3D maps of the road ahead in real time.
The system, which the researchers call MapLite, can allow an autonomous vehicle to form a detailed map extending up to 100 feet ahead. Once it reaches the end of this distance, it creates another map and so on and so forth, until it reaches its destination.
Ort and his team have tested MapLite on rural roads in Massachusetts and have found that it can detect obstacles and curves at that distance. Unlike existing navigational systems, MapLite doesn’t rely on road markings, curbs, medians or other characteristics, instead favoring real-time data of the road ahead and applying only the most basic assumptions about road geography, such as the concept that the road surface will be flatter than the surrounding area.
The team still has a lot of work to do to ensure that MapLite can tackle changes in elevation, weather conditions, low light and the appearance of sudden obstacles in unlighted roads. But the technology is promising because it could bring public transport to small communities that often don’t even have a public bus system or more than one taxi company. For rural residents, especially senior citizens and the disabled, self-navigating autonomous cars could allow them to remain in their homes and communities despite not being able to drive themselves.
Of course, the benefits extend to urban communities, too. Creating detailed and extensive maps of urban roadways is an inefficient, expensive and not very effective way to train driverless cars. MapLite, or a similar technology, could eliminate the need for those maps, solve the many of the current problems with driverless cars and make the technology more accessible to communities of every size.