Researchers at the MIT Computer Science and Artificial Intelligence Laboratory have developed an algorithm to allow autonomous cars make better decisions about changing lanes.
The project employs advanced knowledge of how humans drive and make decisions rather than using the statistical data.
Standard autonomous car collision systems put a predefined buffer around the other cars on the road, so it can avoid crashing into them – there’s always a virtual barrier around them. This is completely turned on its head with MIT’s implementation.
In fact, the system gives the cars less information about how to react and instead varies the buffers at certain points around the other cars to ensure it makes decisions on the fly, rather than the usual, carefully calculated way of thinking.
“The motivation is, ‘What can we do with as little information as possible?'” Alyssa Pierson, one of the researchers explained. “How can we have an autonomous vehicle behave as a human driver might behave?”
MIT thinks this way of driving will ensure autonomous cars have styles of driving rather than all acting the same. The benefit to this? If a car is driving really badly, the buffer zones can be adjusted to deal with that, on the fly, rather than reacting with a standard response.
“The autonomous vehicles were not in direct communication but ran the proposed algorithm in parallel without conflict or collisions,” explains Pierson.
“Each car used a different risk threshold that produced a different driving style, allowing us to create conservative and aggressive drivers. Using the static, precomputed buffer zones would only allow for conservative driving, whereas our dynamic algorithm allows for a broader range of driving styles.”