MIT Made an Autonomous Car Capable of Driving Without 3D Maps
【Summary】Other self-driving cars can’t operate on their own on unmapped country roads. MIT’s new machine is one of the rare exceptions.
To drive on their own, autonomous vehicles need a host of high-tech features. There's the hardware, which includes things like cameras, sensors, and LiDAR. The second part of the equation involves the software, all of the necessary components that do the computing and comprehend what the various hardware parts are seeing.
Maps Are An Integral Part Of An Autonomous Car
One of the major things that autonomous vehicles need to operate properly is a highly-detailed set of 3D maps. These 3D maps provide vehicles with the exact location of lanes, signs, off-ramps, and curbs. Companies and automakers spend thousands of hours and millions of dollars to invest in high-end maps, as they're a crucial part of the equation.
"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," said Daniela Rus, director of Massachusetts Institute of Technology's (MIT) Computer Science and Artificial Intelligence Laboratory (CSAIL). "The need for dense 3-D maps limits the places where self-driving cars can operate."
Incredibly, researchers at MIT's CSAIL have made a system that allows a driverless car to operate on roads without 3D maps. Essentially, MIT's system allows autonomous vehicles to operate on roads they've never seen before and tarmac that doesn't have the best road markings.
MapLite Lets Driverless Cars Drive Anywhere
Called MapLite, MIT's system utilizes simple GPS data that works in tandem with multiple sensors, as well as LiDAR to look at the road. Apparently, that's all it took to have an autonomous Toyota Prius, thanks to the Toyota Research Initiative, navigate unpaved country roads in Devens, Mass. The system reportedly detected road conditions from up to 100 feet away.
"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 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."
As MIT points out, automakers and tech companies may be making a lot of strides with self-driving technology, but they're still not as good as human drivers yet. Their navigation skills are especially lacking when compared to a human driver's. For example, if you need to get somewhere, you probably put the address into your smartphone and place the device into a holder. As you travel on your route, you occasionally glance at the phone.
If the same principle applied to self-driving cars, the machines wouldn't look away from the smartphone for the entire journey, as they rely too heavily on maps at the moment. MapLite gets around this issue with by using sensors for every aspect of navigation. The system relies on GPS data only when it wants to get an idea of where the vehicle is located.
How MapLite Works
MapLite works by setting two points – a final destination and what researchers refer to as a "local navigation goal." The latter has to be within view of the vehicle. Once these two points have been chosen, the system's perception sensors come up with a path to get to the final destination point. LiDAR is used to provide the vehicle with a rough estimate of where the road ends. Apparently, MapLite can operate on roads without any physical markings by assuming the flatness of the road it should be driving on.
"Our minimalistic approach to mapping enables autonomous driving on county roads using local appearance and semantic features as the presence of a parking spot or a side road," says Rus.
Other companies are working on systems that allow their autonomous vehicles to operate without map data, as well. But MapLite is unique because it doesn't rely on machine learning. The majority of companies train their systems on one set of roads and then test them on another. That's not what MIT does.
"At the end of the day we want to be able to ask the car questions like ‘how many roads are merging at this intersection?" said Ort. "By using modeling techniques, if the system doesn't work or is involved in an accident, we can better understand why."
Where MapLite Can Be Improved
As with any high-tech system that's in its infancy, MapLite isn't perfect. The system has its restrictions. MIT doesn't believe the system is good enough to operate on its own on mountain roads. The problem with that kind of terrain, according to the university, is the drastic changes in elevation.
The team, though, is looking to improve upon the system. Eventually, MIT wants MapLite to have a similar level of reliability and performance as 3D map systems, but with a wider range.
"I imagine that the self-driving cars of the future will always make some use of 3-D maps in urban areas," said Ort. "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."
MIT's MapLite system is interesting for a few reasons. Besides not utilizing 3D maps, the system is aimed directly towards consumers in more rural areas. Unfortunately, autonomous vehicles are being marketed towards those in urban locations.
Drivers in cities, at least those that reside in California, aren't really interested in having autonomous vehicles on their streets. While that's a viewpoint that's shared with the majority of drivers, urbanites are more likely to support autonomous vehicles than individuals that live in rural locations. Cities may have horrible, pothole-ridden roads, but it's unlikely that driverless cars operating in an urban environment will operate on unpaved roads.
Still, if MapLite were to find its way to the real world, it would be a more affordable and simplified way of getting autonomous vehicles to operate on their own. And if self-driving vehicles do reach rural areas, the system would result in a better autonomous car.
Vineeth Joel Patel
Joel Patel has been covering all aspects of the automotive industry for four years as an editor and freelance writer for various websites. When it comes to cars, he enjoys covering the merger between technology and cars. In his spare time, Joel likes to watch baseball, work on his car, and try new foods
Electric Vehicle Boom Could Spell Demise for Japan’s Leading Industry
Sacramento Hopes a $44 Million Plan Will Help it Become an EV Leader
Ford’s Plan For Moving Everything Under One Roof is to Draw Investors
British Startup FiveAI Will Start Trials of Autonomous Vehicles in London in 2019
China Quickly Taking Control of Cobalt as it Looks to Advance EV Industry
More Startups Need to Join the Autonomous Car Race, States Fortune
Supply Shortages Risk Bringing Record EV Sales to a Halt
IIHS Find Flaws With Adaptive Cruise Control, Lane Keeping Assist Systems
- BYTON Places Order for Factory to Paint 150,000 Cars Per Year
- Ford in Talks with Volkswagen on a Possible Truck Partnership
- Digital Cockpits Arriving Before Autonomous Cars, Claims Forbes
- GM's Maven Trials P2P Car-sharing Service in Major U.S. Cities
- Texas Startup Edgetensor Develops AI-based Driver Monitoring System for Autonomous Vehicles
- Jaguar Land Rover to Ditch Diesels for Electrified Powertrains
- Tesla to Build China Factory to Produce 500,000 Vehicles Annually
- Didi Investing $1 Billion Into its Auto Service Platform
- Audi, Israeli Autonomous Startup Cognata Partner to Speed up Autonomous Vehicle Development
- GM Seeking Potential IPO for its Self-Driving Unit Cruise