The automotive industry is often at the forefront of technology. It has been using robots on assembly lines for decades. And more recently, it has been among the industrial pioneers in artificial intelligence (AI).
Some of the big problems being faced in the automotive sector right now relate to semi-autonomous and autonomous driving. These vehicles have an array of sensors, cameras, processors, radar systems, and more that provide huge amounts of data in order to avoid obstacles, maneuver in traffic, respond to signs, stay in lanes, come to a stop, park — and everything else required in driving.
Consider just one parameter. It takes about 1.5 seconds to react to an accident on the road. Hitting the breaks requires another 1.5 seconds.
Driverless cars must be able to “see” that there is an actual situation and respond within such a short period. Crunching all the driving-related numbers in real-time requires advanced AI-based processing and decision-making.
6 ways AI is driving automotive innovation
Due to the size of the worldwide market, the profit margins available, and intense competition between the likes of GM, Toyota, BMW, and Tesla, the automotive industry is one of the sectors driving rampant growth in AI.
The total AI market for automotive, including hardware, software, and services, will amount to around $27 billion by 2025, according to Tractica.
Here are some examples of AI in use in automotive:
Nauto has developed a predictive AI alert system that helps you avoid a collision.
It incorporates vision tech used by more than 700 fleets. It processes more than 40 risk factors inside and outside of a vehicle to warn and helps reduce the possibility of collisions by as much as 80%. For example, the Delivery Authority, a last-mile delivery fleet in the greater Chicago area, used it to reduce collisions by 81%.
Nauto combines AI-native technology, data science, and more than 1 billion AI-processed driving miles to predict and help prevent collisions before they occur, according to Yoav Banin, chief product officer at Nauto. Instead of getting lost in vehicle-centric telematics and cameras as indirect proxies for driving risk and reviewing historical events, Nauto’s approach is to use AI to directly understand driver behavior. It analyzes subtle indicators of distraction, drowsiness, cell phone use, and driver attention in the vehicle, combined with vehicle speed, acceleration, and surrounding vehicles and pedestrians to deliver audible alerts only when needed.
“Nauto’s AI is at the edge [in the vehicle] to process driver behavior and external road conditions in real-time,” Banin said. “It is enabling the safe introduction of increasing levels of autonomous driving.”
Tesla is involved in AI on many fronts.
At its recent Tesla AI Day, it just unveiled the DI custom chip as part of Tesla’s Dojo supercomputer system. The chip uses a 7-nanometer manufacturing process and offers 362 teraflops of processing power. With 25 DI chips on one tile and 120 tiles spread across several cabinets, that adds up to more than an exaflop of power — enough to change the AI game in automotive.
Tesla is partnering with Intel, Nvidia, Graphcore, and others on this technology. The goal is to speed the training of AI models to enable them to recognize key specifics from video feeds of Tesla cameras in vehicles. Tesla already uses AI in the chips used in existing vehicles to make decisions within their on-board software, based on what is happening on the road. That enables the company to offer a “full self-driving capability” option for its vehicles to enable them to automatically change lanes, negotiate highways, park, and more.
Kawasaki and SoftBank are using AI to develop next-generation motorcycles that can grow along with the rider and adapt to rider needs.
These bikes use AI for several functions: to provide advice such as slowing down by 5 mph in order to avoid having to stop and arrive at a traffic light when it turns green; notify about surroundings and possible dangers; and advise on road conditions and upcoming hazards, such as steep curves.
Jeep Grand Cherokee is another vehicle introducing advanced AI-based technology.
The latest model has updated active-driving assist to increase driver safety and performance. The company is also using technology from Sight Machine for continuous inspection on the final assembly line for Jeep Grand Cherokees and other cars. This system inspects 1,100 vehicles per day, including 15 exterior elements, and has enough in-build intelligence to differentiate among 25 models and 11 colors with 99.9% accuracy. The data from the inspection is shared between AI and manufacturing execution systems (MES), an image analysis system, and edge/cloud systems.
Sight Machine’s manufacturing productivity platform ”gives every stakeholder from the plant floor to the C-suite a trusted and dynamically updating view of production,” said Jon Sobel, co-founder, CEO, Sight Machine.
“It guides operations with continuous, real-time decision-making. It includes a suite of visualization, data discovery, analytics, and artificial intelligence tools to aid in improving productivity.
Ford is at the forefront of automotive AI research.
To help drive the technology forward, it is harnessing AI in its assembly lines to speed production. At one plant in Michigan, robots assembling torque converters learn how to operate more efficiently via AI based on technology from Symbio Robotics. That’s in addition to having its own drive assist systems and heavy investment in the autonomous vehicle field.
6. Driver monitoring systems (DMS)
A DMS consists of a series of small cameras or sensors placed throughout the vehicle interior that use computer vision (CV) to monitor driver behaviors and issue alerts or warnings when drivers show signs of drowsiness, distraction, or inattention.
These AI-enabled systems can recognize various driver actions: like a driver slumping forward or nodding their head to indicate drowsiness; the direction of their gaze to determine whether they’re watching the road; and the placement of their hands.
“A CV system must account for a huge degree of variance in human appearance, physical orientation, movements, clothing, lighting, objects, and the dimensions and details of cars,” said Gil Elbaz, co-founder and CTO, Datagen.