MIT Teaching Autonomous Cars to See Around Corners
Automakers, companies, and researchers have been trying to come out with cameras that help drivers see what's coming around a corner for years. But few, like Jaguar Land Rover, actually have something viable on the market. A few years ago, the Massachusetts Institute of Technology (MIT) developed a camera system called CornerCameras. Well, researchers from MIT are back with a new system that's meant to help autonomous cars see around corners.
Perfecting Existing Technology
CornerCameras, which came out in 2017, is a system that researchers from MIT came up with that focused on shadows. The system would shine a light into a room from the outside, while sensors and cameras would capture any light that bounced back. Then, it would create a 3D model of objects inside the room. While the tests were completed without cars, the technology was developed with cars in mind.
However, CornerCameras required expensive cameras, lasers, sensors, and all sorts of hardware. As autonomous cars are already incredibly expensive, the system would've made cars even pricier. With price being a major issue for self-driving cars, a system that would drastically increase how much autonomous vehicles cost would be a tough sell for companies.
Now, MIT is back with a new system. In a new approach that will be presented at the International Conference on Intelligent Robots and Systems (IROS) in the near future, researchers at MIT have designed a system that senses and analyzes shadows. According to MIT, researchers claimed that successful experiments have taken place with an autonomous car driving around a parking lot. The team has also used autonomous wheelchairs operating around hallways. Compared to LiDAR, MIT's system can detect objects roughly half a second quicker. LiDAR can only detect visible vehicles, too.
Sensing Shadows To See Around Corners
"For applications where robots are moving around environments with other moving objects or people, our method can give the robot an early warning that somebody is coming around the corner, so the vehicle can slow down, adapt its path, and prepare in advance to avoid a collision," said Daniela Rus, director of the Computer Science and Artificial Intelligence Laboratory.
The system is a modification of something MIT researchers came out with previously called ShadowCam. As its name implies, the system uses "computer-vision techniques" to identify and classify changes to shadows on the ground. It's much more complicated than that, but the gist is that, in MIT's words, the system detects changes in light intensity over time to see if something is moving closer or further away. Depending on what kind of data it receives, the system then classifies each image accordingly and reacts.
All of the testing with ShadowCam has been conducted indoors. While they've been successful, indoor testing isn't nearly as difficult as real-world testing. For one, the robots completing indoor testing aren't traveling at realistic speeds. Then there are the lighting conditions that are more consistent than what you'll find outside.
Still, the results are good and reveal that MIT are on to something that could possibly work on autonomous vehicles in the future.
- Toyota Believes There’s No Market for EVs Yet
- MIT’s Trying to Teach Autonomous Cars to Look Out for Unpredictable Drivers
- Mobileye is Mapping 28,000 Miles of Spanish Roads a Day for Infrastructure Changes
- Daimler Starts Testing Autonomous Mercedes-Benz S-Class Taxis in California
- MIT Study Finds EVs to Cost More Than Regular Cars Until 2030
- Uber, Distracted Driver Cited for Fatal Crash in Arizona Claims NTSB
- GM President Mark Reuss Believes These Three Things Are Holding EVs Back
- Waymo Puts out 14 Minute Video for First Responder Precautions