Articles - Archive

Beyond autonomous driving, the advantages of AI inside a vehicle

From high-tech interiors, new hub for safety and comfort, to engines, this is how AI will affect us even before autonomous driving

Simone Cosimi • Giornalista / Audi Innovative

According to a survey carried out by Intel and Strategy Analytics, once mature, the autonomous car market could be worth the stratospheric figure of seven trillion dollars by 2045. Things are clearly moving in that direction and according to a study by BlackRock (as well as many other experts and observers) in 2025 98% of all circulating vehicles will be somehow connected. A decade later, 75% will be completely autonomous. However, very likely our next car will still not be fully-autonomous. It could possess, though, as in the case of Audi solutions (for example Audi Intelligent Assistance and Audi connect, able to combine comfort, safety and connectivity), a series of tools already deeply affected by artificial intelligence, which calls for us getting to grips with it, inside and around the vehicle, even before “shifting” into first gear.

Obviously, the most fascinating and truly revolutionary angle revolves around driver assistance systems (already largely available) and level 3 autonomous driving or above. Here too, Audi is blazing trails: the most advanced models in the German manufacturer’s range are already able to “read” the road, foresee difficulties and dangers ahead and communicate with the surrounding infrastructures, leading the driver into the future of mobility. However, initially we will likely need to familiarize with a vast array of high-tech wizardry inside the car, a context that, among other things, constitutes a fertile territory not only for the manufacturers themselves but also for a large number of start-ups already active in researching and proposing new and innovative tools.

Among the most promising is facial recognition. However, in this specific case, not used as social networks do but, on the contrary, to prevent accidents. Already present in commercial vehicles and trucks where these devices are busy monitoring drivers to prevent distractions, fatigue or poor judgement, smart cameras have led over the years to lower insurance costs for commercial fleets. Something similar might happen for private citizens and their personal vehicles. iSee, a Boston MIT start-up has been working on making cars more “empathetic”. Still according to Intel, AI could save as many as 500 thousand lives in ten years through accident prevention. In addition, besides warning drivers about their overall mental and physical state (and perhaps, in the case of top-of-the-range models, take over on things such as speed and steering), cameras equipped with alert and recognition algorithms can also be useful for other equally important purposes: alerting the driver if children or animals are "forgotten" in the car. In short, creating a safer and more comfortable environment.

But there is more. Peeking inside a vehicle, careful not to breach privacy rules, can be extremely beneficial even for different systems such as carpooling or ride-sharing. In the first case to assess, perhaps in real time, how many people are in a vehicle and propose a detour to the driver perhaps to load another passenger thus further reducing expenses. In the second case to ensure that all safety procedures are completed before departure, without which, for example, the vehicle would simply refuse to start. Or, in a far simpler, but equally significant way, to warn the drivers about forgetting a bag or some other personal item in a rented car.

Still on the subject of safety, the possibilities are both exciting and endless. Having a “smart-camera” connected to a vehicle’s drive assist systems can play a vital role in an event of an accident. How? For example, by estimating the body mass of the occupants and their position in order to better calibrate and manage the release of the airbags. These issues are little discussed at the moment as the driving side of things take the centre stage, however, the design and technologies commonly found inside our vehicles could undergo a profound transformation. All this without of course considering other elements that touch more on comfort and entertainment such as gesture controls or the possibility of recognizing natural language at a higher level.

The next great metamorphosis will take place under the bonnet. As things stand, a vast assortment of automated systems are already helping us in taking care of our vehicles, reporting malfunctions, or simply warning us it’s time to have the car serviced. In the future, all the data that can already be extracted from sensors scattered around the engine and the vehicle will be sent easily and quickly to the driver. Thanks to machine learning, the “brain” of the vehicle will be able to learn, by reading the routes we normally take and our driving habits, the true meaning of certain information. In the same way, it will be possible to foresee necessary maintenance work without considering mileage as the only viable logic thus preventing more serious mechanical damage. Among other things, these solutions will be increasingly important and will go hand in hand with driverless vehicles: as the interaction with the driver will diminish, a vehicle’s ability to diagnose itself will be increasingly important for the safety of the car. All this without forgetting that, as connected objects and custodians of the lives (and information) of those who occupy them, cars will be the target of increasingly frequent hacker attacks: AI will serve to defend them from these threats, detecting repeated patterns or “cleaning” operating systems from potentially dangerous viruses and malware. Automotive software will therefore need to be intelligent and develop on the wave of recommendations and indications born by different entities, for example the U.S. administration’s Cybersecurity Best Practices for Modern Vehicles released a couple of years ago.

AI will be responsible for a drastic transformation of our vehicles, turning them into some kind of mobile probes and radars thanks to sophisticated cameras and sensor systems which will prove useful not only for their location but also to improve maps and other navigation tools, enriching them with more details in real time, monitoring traffic as no crowdsourcing app can do as well as pedestrian flows and so on.

Obviously most of this is desirable, but it is not certain just how much of it will be fully developed and when. Much will depend on a number of elements, such as technological sustainability, commercial feasibility, acceptance and demand by motorists, and traffic regulations which are often contradictory in different countries, since these connected hubs on wheels, integrate a series of features that go beyond use of an automobile in the strict sense of the word, including privacy, security and research. So much so that in the long run even the interior of a car, a bit like PCs and smartphones, will enter a sort of hardware conflict: manufacturers will have to choose which service to include in their "AI package" (we can call it like that), a further qualifying factor for each model. Including the price, which remains one of the key factors in the choice of a new car.

back to archive