COVID-19 Update: The university remains under a state of emergency. Learn more about guidelines and requirements in place for return to our campuses at Safe and Healthy Buckeyes.

You are here

Our cars understand our driving habits, and that’s valuable information

When the coronavirus lockdowns began in March, delivery services and companies that offer products for delivery saw a surge in demand as people stayed home and ordered everything from groceries to masks to medicines.

Illustration of a delivery truck with multiple data points bursting out of it.

That required many of these businesses to expand almost overnight. It seems like an almost impossible and chaotic endeavor, to hire or increase the use of a bunch of couriers to deliver goods in a satisfactory way.

It turns out, predictive modeling can be an invaluable tool to help these delivery companies expand quickly.

One example came from Mobikit, a Columbus-based data analytics company started by Arnab Nandi, an associate professor of Computer Science and Engineering at Ohio State. Mobikit helped ScriptDrop, which uses couriers to deliver prescriptions to customers’ homes, scale up its company within a month to meet a spike in demand.

Using information from their courier vehicles, software designed by Mobikit gave ScriptDrop “delivery intelligence” or insight into how it could maximize its couriers in the most efficient way possible.

It’s an example of how data analytics can tap into vehicular data to improve a service. Those services could be insurance companies or automakers trying to predict driving behaviors, ride-share businesses scaling up in a given city, and even scooter companies improving the coordination of mobility services.

“Arnab’s research found that organizations were struggling to capitalize on the data that commercial vehicles collect,” said Ritvik Vasudevan ’16, Mobikit’s data solutions engineer. “We realized there was an opportunity to build a software platform around the things telematics and connected vehicle data can enable for these companies.”

It’s the type of machine learning that can help with the evolution of Smart cities or driverless trucks and cars.

How?

It starts with telematics data.

There are three typical ways data is collected from a vehicle, Vasudevan explained.

  • OEM (original equipment manufacturer) telematics: OEM telematics are embedded into the car itself to collect data that can be extracted and analyzed. “It’s the highest quality data, the gold standard because it comes from the vehicle itself,” Vasudevan said. “Many newer vehicles have some kind of connected capabilities, usually expedited over a cellular chip.”
  • OBD (on-board diagnostic) telematics: Originally created to help mechanics plug a device into an OBD port to understand what’s wrong with a car, these can now be used to extract information on how a car is being driven. It can collect everything from how fast a car is going at a certain time to if you’re hitting your brakes when the turn signals are on.
  • Mobile telematics: Yep, your cellphone can collect GPS information to how fast you accelerate at a given time. It’s the lowest quality (data) currently but it’s substantially growing.

“A fleet company may go and purchase OBD ports, or an insurance company may roll out a cellphone app to collect data,” Vasudevan said. “It gives a picture of what’s going on but it can be hard to reconcile that data across different devices or apps to see a unified picture.”

In other words, it’s a lot of data to process.

Going back to our Mobikit example, Vasudevan said the companies he works with – automotive, insurance, fleet and mobility companies – spend 80 to 90 percent of their time cleaning up vehicle data into something useful and the rest of the time actually analyzing it.

Using software to interpret telematics data, Mobikit produces predictive models for these companies to improve operations. For an insurance company, that could be helping them understand what a commercial vehicle driver was doing prior to a car wreck. For the ScriptDrop example, it meant helping them assign couriers based on everything from performance to equitability to high-volume pharmacy needs.

“We exist so companies don’t have to invest hours of data engineering into understanding connected vehicle data,” Vasudevan said. “When a company like ScriptDrop has to move fast, we have a platform built for them to do exactly that.”

These types of partnerships are likely to increase and become more sophisticated, according to Vasudevan.

“You’re now going to see vehicles accommodating three or four businesses. You see people driving for Uber and Lyft, GrubHub and UberEats, but what happens when those vehicles become even more heterogeneous, when they are doing employee shuttling during the mornings and afternoons and grocery delivery during the day. The complexity or way a vehicle will be used going forward is going to become increasingly complex, and that will mean opportunity for us because people will want a fine-grain understanding of how, where and for what intent a vehicle was used.”

originally appeared in Ohio State Insights