Putting more intelligence in truck routing

user-gravatar Headshot
Updated May 31, 2015

How will drivers respond to a computer-generated message of this kind:

Dean Croke, vice president of Omnitracs AnalyticsDean Croke, vice president of Omnitracs Analytics

“You’ve been driving for three hours. The sun will rise in 45 minutes. A parking spot is ahead. Pull over and take a 20-minute nap.”

Many fleets, especially larger ones, already use routing and navigation applications that give drivers turn-by-turn directions and optimized fuel purchase plans.

Will drivers follow more intelligent route calculations, like the example above, to reduce fatigue?

A few years ago, changing routes based on predictive modeling might have seemed more like high-risk gambling than intelligence. Navigation and routing applications already got drivers from point A to B. As for when and where to sleep, that decision was up to the individual.

Electronic logs have not made that decision any easier for drivers. Without the flexibility of paper logbooks or rules that allow a split sleeper berth, electronic logs merely enforce compliance rather than give drivers useful information to manage fatigue.

“You can be 100 percent compliant with the hours of service but be sound asleep at the wheel,” says Dean Croke, who has made a career out of unraveling the complexities of driver fatigue and truck accidents by using predictive models and lots of data.

He didn’t stumble into this field by chance. His career began by following his father into the trucking industry where he drove for more than two million miles as a trucker or “truckie” in Australia.

As a driver, he experienced falling asleep behind the wheel hauling fuel and cattle through the Outback. Those days are behind him, but he still owns a truck, a 379 extended hood Peterbilt, which he uses to haul freight on the weekends along I-95 in the Northeast.

Croke is now working on intelligent routing as the vice president of Omnitracs Analytics. The company is developing a routing and navigation system that incorporates predictive models to identify at-risk drivers for fatigue and accidents.

For instance, suppose a fleet has a route that goes into New York City. Eventually, Croke says the software could notify fleet managers if the driver assigned to the route today is more likely than other drivers in the fleet to wreck. With this advance notice, it would probably be wise to assign the route to someone else.

Building up

Croke was one of the founders of FleetRisk Advisors, a company that developed predictive models for fatigue, accident frequency, workers comp claims and driver turnover.

When the company began 10 years ago, nothing like this had ever been done in the trucking industry. When the company landed its first customer, Dupre Logistics in Lafayette, La., “we weren’t sure we could do this,” Croke says. “We had no idea that we could use data to predict anything.”

The debut was successful and FleetRisk began working with other major carriers like C.R. England, Covenant, Maverick, Averitt Express, Schneider and Swift. The company became known as Omnitracs Analytics following an acquisition by Qualcomm that led to the acquisition of its Omnitracs business unit by Vista Equity Partners in November, 2013.

Its fatigue model uses sleep science and electronic log data to accurately predict when and how much sleep drivers obtained in a 24-hour period. By using its Web portal or integration to third-party dispatch systems, fleet managers can monitor for exceptions like drivers who have been awake for 16 continuous hours.

In terms of reaction time, this level of fatigue is equivalent to driving drunk with a .08 blood alcohol content. Drivers that have been awake for 24 hours or more have the equivalent reaction time of .10 BAC, he says.

Recent developments

Predictive analytics has changed a lot since Omnitracs Analytics built the first models. The amount of data has increased exponentially along with the speed, power and intelligence of its data processing. To put it in perspective, the company now processes 29 Gigabytes of data every day.

The models are tailored to each customer and use thousands of data points to identify at-risk drivers and the recommended actions to prevent a likely event such as an accident or voluntary termination of employment.

In most cases, the recommended action is a conversation with the driver to alleviate the stress, anxiety or frustration from their personal or professional lives that is causing a change in behavior and attitude.

Omnitracs Analytics is monitoring text messages of drivers and fleet managers to identify patterns that can be used to predict turnover and accidents.Omnitracs Analytics is monitoring text messages of drivers and fleet managers to identify patterns that can be used to predict turnover and accidents.

Croke shared a few details about the intelligent routing application that is currently being developed during the CCJ Spring Symposium on May 20-22 in Birmingham, Ala. The company recently hired Rick Turek, the architect of the Maptuit NaviGo application (acquired by Telogis) to lead this effort.

During the CCJ Spring Symposium, Croke mentioned other ongoing projects and insights gained from predictive modeling. They include:

  • Text messages: Omnitracs Analytics has been analyzing text messages sent by drivers and driver managers through the Omnitracs in-cab MCP platform. “The data is unstructured but when you look at 134 million words a month patterns emerge,” he says. These patterns have found that certain words and frequency of messages are highly predictive of turnover and critical safety events.

Patterns in text messages can be combined with patterns found in other data sets such as the number of miles driven or a change in deadhead miles to predict turnover and accidents.

  • DOT physicals: The timing of DOT physicals is another leading predictor for accidents and turnover. As drivers approach the expiration of their medical card they are likely to become stressed. The highest turnover happens at nine months before a DOT physical and the lowest turnover rate is 16 months before. There is a sudden spike in turnover and accidents immediately after getting a card, especially for drivers who get a short-term medical card.

Croke believes the turnover is due to drivers choosing to leave the industry rather than make lifestyle changes like wearing a CPAP device to correct sleep apnea problems.

“There is something about a DOT physical that changes behavior,” he says.

  • Dwell time at docks is another useful predictor of turnover and accidents. Turnover is the highest among drivers that are detained at shippers between 72 and 78 minutes per 28-day period. It is lowest among drivers between 54 and 57 minutes.

“As minutes go up on docks, the chances increase for a critical event because drivers are rushing to do stuff,” he says.

  • Team pairing: Omnitracs Analytics is working with one of its customers to determine the best way to pair team drivers by using sleep patterns, personality types, etc.

As the amount of data and processing power continues to accelerate, so do the opportunities for using predictive models and mobile technology to present the right information to the right people at the right time to avoid costly and preventable mistakes.