GM has filed a patent application for a system that can detect if someone is impaired by analyzing their gait. GM’s patent application is designated US patent number 2026/0062025 A1 and was filed with the United States Patent and Trademark Office (USPTO) on September 5th2024. Submission published March 5th2026, and lists four US-based engineers as its inventors, including Siddhartha Gupta, Senthil Karuppaswamy, Mitesh Lalwala, and Alok Warey.

Essentially, the system described in GM’s patent is designed to analyze the way a person walks to determine whether they are impaired – for example due to drug/alcohol intoxication, or whether they are experiencing a medical problem. The system works by using cameras and sensors to evaluate a person as they approach a vehicle. As a person walks toward a vehicle, the system collects data regarding his gait, including his walking speed, stride length, whether he sways to the side or not, and whether he walks in a straight line.
This data is processed using machine learning systems, including Long Short-Term Memory (LSTM) recurrent neural networks. The system analyzes data and images to determine a “gait score,” which is used to represent a person’s likelihood of experiencing impairment. The model can also incorporate other analysis techniques and methods to evaluate an individual’s gait.

Once calculated, the gait score is compared with a predetermined threshold, and if the score does not exceed the threshold, the system allows the vehicle to be operated normally. If the score indicates potential interference, the system can respond in several ways, including notifying individuals via visual or audio alerts, activating secondary detection systems such as breathalyzers, or even restricting vehicle operation completely.

The system can also confirm that an approaching person is indeed the driver by detecting things such as if the driver’s door is opened or an attempt is made to start the vehicle. In addition, this system incorporates methods to differentiate between different individuals and can learn certain patterns associated with impaired and non-disordered behavior.


