Artificial intelligence: Doing more than just keeping your car in its lane

Everyone has heard that creak or noise coming from under the bonnet, only to be stopped when it’s too late and the car is steaming and sitting on the side of the motorway. A survey by UK car repair company Kwik-fit shows that the average cost of a breakdown repair is around £350, but in some cases, the cost can be over £1,500. Many of the faults that require repair could have been spotted before reaching the point of failure, but there isn’t always a mechanic in the passenger seat, so how can we predict these problems and deal with them as early as possible? The answer is artificial intelligence (AI).

AI is not limited to driver assistance

From Super Cruise to Autopilot and beyond, the field of Advanced Driver Assistance Systems (ADAS) and fully autonomous driving has become a mainstay of the automotive industry. Many of the latest and most advanced AI systems are ready to go in automakers’ most premium vehicles. AI can do more than just help keep a vehicle in its lane, seamlessly back into a parking space, and direct you to your favorite coffee shop; AI is very good at recognizing patterns and predicting outcomes based on simple sensor data that may be too granular for human senses to perceive, making it the perfect tool to catch unusual noises before they become “expensive” bangs.

AI needs to be trained with high-quality data to work effectively. Fortunately, vehicle failures, repairs, and maintenance are well documented, while failures recorded by mechanics and annual tests performed by transportation departments and others can provide a wealth of empirical evidence of vehicle operation. Taking this data and feeding it into a machine learning algorithm can provide an efficient model that can predict possible failures based on a variety of data points ranging from exhaust emissions to slightly uneven brake fade.

We already have an all-knowing mechanic in the cloud, and now we need to bring that intelligence to the vehicle at the edge. The next step is to connect a very basic neural network accelerator (like Imagination’s NNA) to the vehicle’s engine management computer and other existing sensors in the vehicle, such as tire pressure and temperature sensors. With a few additional sensors that can measure things like abnormal vibrations, AI can spot potential failures before they happen.

That way, when the AI ​​detects a condition that it thinks may soon turn into a fault, it alerts the driver to what has been detected and recommends that they go to a mechanic for a quick, inexpensive repair, rather than sitting back and eventually getting pulled over. The system can also provide more detailed information about the perceived fault to the mechanic, allowing them to spend less time and effort on troubleshooting, freeing up their energy to provide a high-quality repair.

It may sound like a lot of money to put this kind of “AI mechanic” in every car, from the most luxurious model to the most basic short-distance city car, but when a breakdown occurs, it’s not just the customer’s pocket that suffers.

Minimize the costs of failures

Frustration and anger are two of the most common emotions when a breakdown occurs, some of which are specific to the circumstances of the breakdown, but often these emotions are directed towards the vehicle and its manufacturer. In addition to the hundreds of pounds it will cost the driver to repair the car, if the driver seeks to buy a new car, it can be a huge loss to the original car brand, as he will swear off buying your brand again, meaning your sales pitch will be of no use to him.

There is also the question of who is actually paying for the repairs. Vehicle ownership has changed dramatically, with more people than ever before opting for leases, financing or loan-based ownership models. These ownership models often include extensive warranties and repair plans, in some cases for up to 100,000 miles and 10 years.

In the eyes of consumers, a long warranty means that everything is taken care of, and if something goes wrong, they naturally assume that the dealer or automaker should cover the cost of repairs.

If AI systems were used to monitor the status and potential for failure of millions of vehicles, manufacturers could detect and proactively address potential system problems before they became larger issues.

Fast pace of life

No one likes problems, and car breakdowns are no exception, whether for individual car owners or fleet managers who worry about the impact of repairs and fleet downtime. By deploying simple, low-cost neural network accelerators and other computing devices, many of these headaches can be solved before they occur, turning the smoke and driver anger of a car breakdown into a relaxing trip to the garage, where all that’s needed is a simple repair under warranty to keep the customer on the road.

Imagination has the industry-leading Neural Network Accelerator (NNA) IP series, which are dedicated AI accelerators that can provide leading edge AI inference performance. Since the release of 2NX NNA in 2017, Imagination has released three generations of NNA products. Among them, 3NX can help automotive applications optimize computing power and performance; and the latest generation 4NX is the ultimate AI acceleration solution specifically designed for ADAS and autonomous driving applications. In addition to providing ultra-high performance and ultra-low latency, in terms of automotive safety, it can also help customers obtain ISO 26262 certification faster with a design process that complies with the ISO 26262 standard.