Automatic driving and game theory
https://www.wired.com/story/why-tesla-designing-chips-train-self-driving-tech/
Recently, I read a news which talks about why Tesla is designing its own chips to better empower then automatic drive function of the electronic cars. Elon Musk said the new chips would be used to train company’s neural network, which is the key to a safer and more efficient auto-drive system. Taking the advantage of large costumer base, Tesla gathers more user data than any other companies in the industry. By continuing improving the algorithm, automatic driving cars would be able to handle more sophisticated traffics and make the decision on the road just like human beings. There are so many decisions to make on the road, for example, when to accelerate, when to slow down, whether stop in front of the crossroad, whether keep away from bicycles that have in intention to across the street. These are all the decisions our drivers have to make when driving, so do the AI in drive-less cars.
The decisions drivers have to make I mentioned above is related to what we learned in class, which is the game theory. Payoff matrix is everywhere in the traffic, and driver’s decision could largely affect the driving efficiency and safety. For example, in a crossroad where there is no traffic light, there is a payoff matrix when two cars(one goes from south to north, another goes from east to west) meet together. Each driver will have to decide whether go first or let the other car goes first. If one driver choose to go first and the other driver decide to wait, then the traffic would be the most efficiency. If both drivers choose to go hurriedly, then non of them will across the road until one driver finally let another car to go first. In this circumstance, both drivers waste their time and wait longer than the previous circumstance. Even worth, if there are more cars follow behind them, the crossroad might be crowded and no one could pass through. Last but not least, if both of cars wait behind the line, they would both waste their time although the risk of causing traffic jam is reduced.
This is a simple example on how the game theory could be applied into the traffic, there are more of that, and they are the reasons why automatic drive car companies like Tesla invests on the chips, gathers user data, and keeps improving their neural network algorithm. If in the future, those drive-less cars could always make right decision when facing to the complex traffic that applies game theory, the traffic will definitely be more efficient.