Apple wants to introduce machine learning into apple cars: to make driving decisions quickly

Apple is reportedly planning to use "machine learning" (ML) technology in its "Apple Car" electric car because the current processor is not fast enough to automatically make some critical driving decisions.

In fact, it was already expected that Apple would introduce machine learning into the Apple Car, especially after John Giannandrea, head of artificial intelligence (AI) and Siri, took charge of apple car development.

But now, a newly disclosed patented technology further confirms that speculation. In this patent document, Apple explains how machine learning technology can be applied to the Apple Car and why the Apple Car needs it.

It's all about the speed with which decisions are made. Even a correct decision, such as changing lanes or avoiding a collision, can be fatal if it is not done quickly enough.

Apple said in the patent document: "Until recently, due to existing hardware and software limitations, the maximum computational speed used to analyze the external environment of the vehicle was not enough to make important navigation decisions without human guidance." ”

Although current computing hardware and software are getting better, Apple still believes that this is still not enough. Apple said: "Even with high-speed processors, large capacity memory and advanced algorithms, making timely and reasonable decisions about the vehicle environment is still a huge challenge." ”

Apple also spoke about the complexity of autonomous decision-making, saying it was based neither on overly pessimistic or overly optimistic assumptions. Cars may be able to drive themselves, but they will never drive alone. So, one factor to consider is the "unpredictable behavior" of other drivers in other cars.

In addition, the real world is much more chaotic than any test environment, so Apple also points out that in the case of insufficient data completion, the decision to drive itself should also be made. This also reflects the complexity of autonomous driving.

Apple also said that in some states, the number of actions to be evaluated may be relatively small when vehicles are driving on a nearly empty straight highway where it is impossible to turn for kilometers or miles. But in some other states, the number of actions can be much larger when vehicles approach congested intersections.

In each case, the car's system must determine the "current state of the environment" around the vehicle. Then, determine "a set of corresponding actions that can be taken." An action can be "turn left" or "change lanes". At least in some cases, machine learning can be used to help cars assign a number or value to each possible decision and then determine the best course of action.

Source: Sina Technology