In what's believed to be a first in the automotive industry, Mitsubishi Electric is using machine learning algorithms to analyze vehicle data (like speed and steering) and driver behavior (heart rate and the orientation of their head) to predict possible actions. The technology constantly compares "normal driving" to its algorithmic prediction of what is appropriate, using data generated from previous journeys, and immediately alerts the driver of their potentially reckless actions.
Because Mitsubishi's machine learning technology processes forward- and backward-moving data, the system has already reduced driver prediction errors by 66 percent compared to its old algorithm, a three-layer neural network. The company will show off its new technology at the Tokyo Motor Show later this month and could be integrated in a new wave of driver-sensing units installed in new cars from 2019 and onwards.
[Image credit: drs1ump, Flickr]