The Future of Autonomous Driving with Artificial Intelligence

The automotive industry has become especially ripe for digital disruption after NVIDIA announced the world’s first artificial intelligence computer to support fully autonomous vehicles in October 2017. This date marked the proliferation of artificial intelligence (AI) and brought the wave of seismic changes to the automotive industry for the last year.

Current drivers already admit that they would spend an extra $500 to $2,500 per vehicle for various advanced driver-assistance features, according to McKinsey. The researchers anticipate that around 50 percent of passenger vehicles sold in 2030 will be highly autonomous and approximately 15 percent will fully take over control and execution from drivers. It’s difficult to imagine that in a decade, roughly every seventh car in the United States is expected to run on its own in a driverless condition thanks to AI.

What is AI in Autonomous Driving?

In simple terms, artificial intelligence is the ability of a machine to think logically, learn, and make decisions. To distinguish signs from pedestrians, or perform any action on the road without the involvement of drivers, autonomous vehicles use a complex combination of AI processing units mainly based on deep neural networks. Feeding computers with loads of data, we rely on them to analyze, process, and perform actions just like human beings. We call them electronic brains and expect them to function like our own to eventually exceed our intelligence.

In the context of automotive, AI is an umbrella term that incorporates the vehicle’s computer vision, deep learning, and decision-making capabilities. The Society of Automotive Engineers (SAE) has established a standard with five levels of cars, where level-4 and level-5 are regarded as highly and fully autonomous vehicles. Level-4 allows drivers to run a car with their minds off. In the cars of this type, self-driving is possible only in geofenced areas or under a limited set of conditions. In turn, level-5 neither requires a steering wheel nor driver’s intervention, as it fully cedes control and governance to the car. To make it happen, manufacturers apply myriads of hardware and software components that enable a robotic vehicle to safely navigate across the streets on its own without any threat to the environment.

AI-based systems are fundamental in the automotive industry as they form the core of Advanced Driver Assistance Systems (ADAS) and contribute to the infotainment human-machine interface. The year of 2018 has been a successful reset season for the automotive game-changers, especially in these categories. AI already helps the infotainment systems inside the cars recognize speech and gestures, perform eye-tracking and virtual assistance. The cars can now learn from shared experience and make accurate predictions and suggestions.

AI to Enable Cars to Mimic Human Cognition

But fully autonomous vehicles have to learn to contend with all other factors. The main challenge out there in front of autonomous vehicle manufacturers, though, is to process data collected from numerous sources, such as cameras, LIDARs, GPS, ultrasonic sensors, and many more. The turning point would be to provide autonomous transport with cognitive and intuitive capabilities and make sure that vehicles of the new generation can think and reach decisions as drivers would normally do.

Manufacturers should boil the ocean to ensure that their cars react in ambiguous situations and consider all “if-then” scenarios that might impact the course of driving. As this goal seemed infeasible in terms of time, location, and resources, carmakers began to look for a more optimized way of building solutions — creating a connected fleet of cars that could learn from each other. In a fast-paced learning environment, cars can reach the highest level of self-governance — drivers will be able to put their hands, eyes, and minds off when on the road.

AI to Replace Steering Wheels

Even though many manufacturers have started their journeys to computers on wheels, there are only a few companies in the world at the forefront that have blazed a trail to fully driverless experience. Waymo, Google’s subsidiary, has been developing in the industry since 2009 and today it shows the highest level of autonomy among the cars that handle the driving.

Even though you can spot a Waymo car having a steering wheel and pedals, the initial prototypes of cars presented by Google in 2014 were designed as free from ones, but manual controls were fitted in to meet street legal requirements. Waymo’s ambitions are high, with 8 million miles recorded on public roads and over 5 billion miles in simulation, the company strives to beat the competition.

Google’s former car, however, isn’t alone at this mountaintop. General Motors’ 2018 Self-Driving Safety Report has just presented a new zero-emission self-driving vehicle of level-4 — the Cruise AV with the highest levels of automation in perception, planning, and control processes in the world. The company reports that this experience paves General Motors the way to the level-5 car in 2019.

AI to Navigate Uncharted Territories

Still, there are a few missing pieces of the puzzle that leading manufacturers have to consider to make fully autonomous cars. In May 2018, MIT researchers made a great leap forward to driverless cars by presenting an autonomous vehicle that can navigate unmapped roads. The developed system, called MapLite, does not require 3D maps, as it relies on simple GPS data combined with multiple sensors that observe the road conditions. The system can be of great help to communities that live in rural areas.

AI to Add Fuel to Moral Dilemma

The major bottleneck complicating the release of autonomous cars on the streets without drivers is a moral dilemma or controversial situations that imply life-threatening outcomes to the individuals involved. The autonomy given to self-driving cars will sooner or later result in unpredictable accidents and push the computer to make choices. What factors will drive AI to take directions in the critical momentum? Who is to blame?

The goal of producing next-generation cars would never become achievable without the powerhouse of AI-driven algorithms and top-notch project management skills. The role of AI in the software is imperative. By analyzing volumes of information that comes from different kinds of sensors attached to cars and arriving at sensible prompt decisions that would run a vehicle, AI translates the knowledge into concrete steps and thus sustains the next generation of autonomous transport.

Iryna is a Founder & Chief Editor at PM Column, a creative magazine for digital project managers. She is passionate about management and technology when they go together.