GM Patents Autonomous Driving Instructor


General Motors has filed a patent for a driver-training system that utilizes a vehicle’s onboard sensing equipment to determine how well a novice motorist is handling themselves behind the wheel.

The objective is to offer driver education without the help of a flesh-and-blood instructor being present. Instead, the autonomous vehicle limits the amount of control offered to the student while constantly monitoring their progress. If they score well enough, additional freedom is awarded to the driver and the process begins again — this time with the vehicle looking to evaluate more advanced maneuvers while still keeping tabs on the basics. It’s quite a bit different than the standard practice of having someone sit beside you to take stock of your budding driving skills. But GM thinks it might have future applications and probably wants to lock it in with the United States Patent and Trademark Office (USPTO) before anybody else does. 

This is supported by the highly generalized language used in the filing, first shared by Motor1. While we’ve no doubt that the company (and probably a few others) have been working on the concept of a computer-controlled driving instructor, there’s not much to the abstract. There’s also some word soup in the summary with plenty of wiggle room for implementation. The filing casts a wide net for how this system would actually work, likely as a way to keep itself covered in terms of design variation. But that’s fairly standard for USPTO filings.

From the abstract:

In accordance with an exemplary embodiment, a method is provided for training a trainee using an autonomous vehicle, the method including; measuring, via one or more sensors, one or more manual inputs from the trainee with respect to controlling the autonomous vehicle; determining, via a processor using an autonomous driving algorithm stored in a memory of the autonomous vehicle, one or more recommended actions for the autonomous vehicle; comparing, via the processor, the one or more manual inputs from the trainee with one or more recommended actions for that autonomous vehicle, generating a comparison; and determining, via the processor, a score for the trainee based on the comparison between the one or more manual inputs from the trainee with the one or more recommended actions for the autonomous vehicle.

It’s almost tragic that the people working at the patent office have to read stuff like this all day and I’m genuinely sorry you just had to spend a moment in their shoes. But the gist is that GM has come up with a concept and would like to retain control before one of its rivals come up with something similar.

The automaker says the assumed benefits include the elimination of any instructor bias and reduced cost of existing training programs. Theoretically, this could be a feature General Motors’ simply bakes into future products (assuming self-driving cars ever go mainstream), sells via an over-the-air update, or loans out to driving schools. As a byproduct, the system could also amass value driving data for the company. As for the hardware required, GM left things pretty vague. However the paper does make reference to camera arrays, lidar, and global positioning systems (GPS) that have been integrated into vehicles as potential solutions. Other than that, cars would need to have an onboard processor to evaluate how well the trainee is progressing and keep its own self-driving capabilities on task.

But the manufacturer did admit that there could be a few blind spots, most notably the possibility that a human driver would need to assume complete control of an autonomous vehicle on occasion. This includes instances where some of the sensing equipment becomes damaged or obscured and whenever the car finds itself on an unmapped road it doesn’t know how to navigate.

[Image: Demskoy Studio/Shutterstock]

Become a TTAC insider. Get the latest news, features, TTAC takes, and everything else that gets to the truth about cars first by subscribing to our newsletter.

Source link