News

MIT’s New Algorithm Teaches Self-Driving Cars To Understand Their Owner

It was only a few months ago, Elon Musk was predicting the arrival of autonomous Tesla cars, and here we are with thousands of self-driving vehicles on roads as we speak.

But the system still needs to get past a number of challenges. One of the biggest ones appears to be the relation between humans drivers and AI. Humans are selfish, cautious, and overly unpredictable on the road, and machine learning still needs to wrap its head around that.

MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) researchers have found a solution to his problem. They have created an algorithm that can predict the behavior of drivers with respect to impatience and selfishness.

In the real-world tests, the MIT researchers observed 25% greater accuracy in the behavior prediction of the other cars. In other words, autonomous vehicles were able to make sound judgments with the new algorithm.

Behind the hood, the researchers are using the Social Value Orientation of social psychology. It essentially prioritizes a driver over others by analyzing their social behavior in an external environment.

In order to make this work, the algorithm initially analyses the driver’s most effective responses on the road. Here, the system looks at how much a driver weighs their own benefit against the benefit of another driver, and weighs it under the Social Value Orientation.

The algorithm categories the driver’s behavior and all the drivers in the surroundings and make intelligent decisions while on the road.

The researchers are continuously working on the new algorithm in order to make it ready for real-world driving. Also, researchers will try to apply the model to pedestrians, cyclists, and several road users.

To Top

Pin It on Pinterest

Share This