Coventry University in the United Kingdom recently took the wraps off its self-driving vehicle, the Nissan ENV 200, an electric-powered vehicle, developed through the Centre for Connected and Autonomous Automotive Research.

The purpose of the vehicle is to explore various aspects of self-driving technology, including connectivity, cybersecurity and automated testing.

The vehicle features a 360-degree scanning LiDAR, cameras for localization, a forward-facing radar, ultrasound sensors and a 5G development kit, and it is intended to serve as a prototype for a self-driving solution for future vehicles.

Autonomous vehicles (AVs) operate in a very complex, multi-variable and dynamic environment and need to be able to rapidly sense and respond to changing driving conditions and events in real time. AVs need to be able to handle variables such as: Vehicle performance, traffic, navigation, road conditions, construction, weather, pedestrians, bicycles, emergency vehicles, traffic signals and unexpected actions by other vehicles; a challenge increasing exponentially, relative to the level of autonomy.

“Currently, AI is the only technology that allows AVs to recognize objects in their environment and respond accordingly,” says Peter Maithel, automotive industry principal for Infor. “In the real world, inputs come in a variety of forms, and these are often mixed in with each other, as well as other data.”

They may also not be clearly defined – for example, when driving in rain or fog, the object in front may only be visible as a vague shape. This shape could be a vehicle, or it could be an obstacle – which would require different responses from the AV.

“AI can be taught to handle these situations, to increase the level of autonomy while also maintaining appropriate safety,” Maithel says.

He notes several automakers are deploying various levels of AI-based Automated Driver Assistance Systems (ADAS) in their vehicles.

Mercedes plans to introduce Drive Pilot, its level-3 autonomous drive technology, in the US. Additionally, Tesla already offers a limited self-driving mode, as does GM with its Super Cruise system.

Pedro Pacheco, analyst at research firm Gartner, explains there are two main reasons that AI is required for the development of autonomous vehicles.

“One is in terms of perception, meaning you have cars equipped with sensors all around and the digital brain of the vehicle needs to make sense of that data,” he says. “You have radar, cameras, lidar all collecting data.”

There then needs to be a conclusion extracted from that data, which is where AI will help determine the operation of the autonomous vehicle.

“The second point is how the vehicle interprets the information of the world around it and decides what it’s going to do next in relation to that situational awareness,” Pedro adds. “This is obviously a key area where you need AI, because the vehicle needs to decide what to do. These are very, very complex things to achieve.”

Maithel adds as vehicles become more software-driven, the role of the actual vehicle hardware in the user experience will shift.

“AI technology will become ever more crucial to delivering a differentiated customer experience, as increasingly, users will interact with software-driven features to create a personalized experience,” he says.

From Maithel’s perspective, automakers who don’t offer these features run the risk of losing to their competitors who do.

“Without the use of AI, automakers run the risk of developing and manufacturing products that are out of sync with customer expectations, thereby negatively affecting customer experience, brand loyalty and retention, which ultimately will affect sales and financial performance,” he says.

Outside of the vehicle itself, automakers who don’t effectively use AI run the risk of falling behind on innovation and missing new business models and revenue opportunities. Pacheco says he thinks higher learning institutions like Coventry and others are particularly important for the research and development of AV systems and the integration of AI because the current limitation of the latter is why self-driving vehicles are not more advanced.

“These institutions play a major role in terms of innovating in the area,” Pacheco says. “Autonomous vehicles don’t learn by themselves. There’s a lot of heavy lifting in which you need to collect a lot of data.”

That’s followed by the building of simulations on which to train the AVs, establishing an environment for how to deal with an enormous number of different situations.

“Universities have a massive role to play when it comes to advancements in AI, because there’s still a lot missing in this space,” he says.