AI Uses Personalization for Accurate Range Estimates

A UK-based startup claims to have developed an in vehicle journey prediction system that uses live data from driver efficiency, weather and route and can estimate battery range more accurately as a result.

Newmarket-based Spark EV Technology has developed machine learning-based technology which it says can predict range more accurately than many of the in-car systems in the latest electric vehicles. TU-Automotive spoke to Spark CEO Justin Ott to learn more about how the machine learning works and why’s so much more accurate than other range estimators.

“Current onboard systems don’t know the individual driving the vehicle. So instead, we’re making a personalized prediction of energy, and we’re saying whether or not you can make your journey. And at the end of the journey, we are comparing predicted energy against actual energy, and then we are refining those predictions,” said Ott.

Spark is licensed to OEMs as embedded software, meaning there is nothing to plug into the OBDII port as is usual with this type of technology. Ott said the company is working with leading automotive OEMs, to integrate the technology directly into a car’s infotainment system, with one manufacturer aiming to release the first vehicle featuring Spark’s technology in the first quarter of next year.

“When you try a new EV for the first time, when you go into the showroom, a car salesman can set up a demo where when you get into the car, input your name, and it will start collecting and learning about how you drive,” Ott told me.

However, it may take a little while for the system to gain that accuracy, as it needs sufficient data to make its estimations. “After around 30 to 50 journeys, the driver will get very accurate range estimations based on their personal usage,” Ott said after I asked how long it would take for the system to be accurate. The technology has completed over 10,000km of trials, with both Hyundai and three other OEMs across Europe and US.

While Ott indicated owners would have to tell the car who was driving, after that, the system wouldn’t need any extra information inputted in order to make the estimations, with no cutting edge – and expensive – biometrics required.

Ott went on to say that many EV drivers have very different ways of driving, and this can have a large effect on range estimations. “What we’re looking at is how efficient or inefficient a driver is, how they’re taking corners, and how they’re looking after their vehicle. The individual driver has a surprisingly high impact on how much energy is used, which means that the personalization piece is really important if you’re sharing this vehicle, with family, or you lease your car.”

Leave a comment

Your email address will not be published. Required fields are marked *