Can AI and analytics make a difference for automotive?

By using advanced analytics techniques including AI, machine and deep learning, as well as computer vision and optimisation, innovative automotive companies are actively pursuing improvements in object segmentation and tracking, sensor calibration, localisation, vehicle control, and path planning. All of these technologies are helping the automotive industry to build an analytics-empowered approach to testing based on an ever growing data set.

Where is the data coming from? As connected cars increasingly stream information into the cloud from telematics systems, infotainment systems, and the array of smart IoT sensors, each connected vehicle, according to a McKinsey estimate, is apt to produce more than 25GB per hour. But how is this data being used and how can the automotive industry best leverage the power of AI, machine learning and analytics to remain competitive?

Driving experience:
Machine Learning applied to connected cars will create a unique picture of driver behaviour. Some cars are already on the road with the capability to remember driver preferences for settings like seat position, internal temperature and favourite radio stations. Based on ever improving machine learning techniques, customisation and automation features like these are only likely to increase. This information will eventually be used to generate each driver’s unique in-car environment, automatically adjusting according to whoever gets into the driver’s seat.

A Massachusetts Institute of Technology spin-off called iSee is doing just that. Instead of relying on simple rules or machine learning algorithms to train cars to drive, the startup is taking inspiration from cognitive science to give machines common sense and the ability to quickly deal with new situations. It is developing algorithms that try to match the way humans understand and learn about the physical world. The aim is to develop self-driving vehicles that are much better equipped to deal with unfamiliar scenes and complex interactions on the road.

Collision avoidance:
Through the use of advanced sensors, big data and car-to-car connectivity, predictive analytics models and AI are working to make car accidents a thing of the past. As developers create applications that increase communications between connected vehicles, more complex and more effective collision avoidance systems will emerge based on predicting drivers’ behaviours through the use of advanced analytics. Of course, the more machine learning that is applied, the more intuitive and understanding the cars will become. Volvo's Drive Me autonomous caris capable of handling any situation that it comes across without any human intervention. If something goes wrong, the car can safely stop itself at the side of the road.

Maintenance:
Predictive maintenance analytics applications can pull in data from virtually every vehicle of a given year and model and compare that information with warranty repair trends. As predictive analytics gains access to increasingly larger datasets, manufacturers can use the information to save both customers and manufacturers money in the long run by preventing the need for emergency repairs or in the case of the manufacturers, the need for expensive recalls. 

Security:
With Gartner forecasting that a quarter billion connected vehicles will be on the road by 2020, it’s easy to see why security is a key issue keeping industry leaders awake at night. The fact is, connected vehicles are no less susceptible to cyber attacks than any other device with an internet connection, but the consequences of a security breach could be infinitely more catastrophic, putting human lives in danger.

What makes predictive analytics effective at securing connected cars, where conventional security measures might fail, is its ability to identify patterns. At some point, every intruder’s behaviour will differ from that of an authorised user. Predictive analysis doesn’t just look for an intruder to repeat the same behaviours as previous attackers. Instead, it looks for any behaviour or combination of behaviours that are not consistent with what would be expected of an authorised user.

This was highlighted two years ago when two security researchers, remotely hacked a Jeep Cherokee via its internet connection to paralyse it on a main road. However, since then, the same researchers have been working with businesses like Uber to help secure experimental self-driving cars against exactly the sort of attack they proved was possible on a traditional one.

Attracting customers:
Traditional marketing strategies have kept most car makers afloat, but today every pound spent on advertising must go further than ever. By tapping the increased availability of data, analytics applications can accurately identify persons most likely to purchase a vehicle in the near future. Complex algorithms consider such factors as time till last purchase, a number of repairs on current vehicle, current mileage, the percentage of the last vehicle paid off before trade-in, and information culled from social media to identify likely buyers.

There are already products doing this, like the Verizon Hum, which is a two-piece hardware system, combined with a phone app that highlights shows full descriptions of any error codes happening inside a car, along with price ranges for fixing them.

Analysis of tomorrow
Connected cars are the biggest innovation in automotive technology in the last few decades, so it’s expected that we hit a few bumps in the road before it gets off the ground. Once it does get moving, the connected car and advanced analytics will help reduce traffic accidents and fatalities, as well as provide a solution to the high levels of traffic congestion in heavily populated areas. This is, of course, all speculation but based on the impact that predictive data has had thus far on the automotive industry; we’re excited to see what comes next.