Predictive analytics are being employed in interesting new ways to improve safety. Telematics solutions have long monitored events like hard braking and speeding to flag unsafe driver behaviors. But today, this driver event data is being enriched with other data streams to actually predict the likelihood of a specific driver having an accident.
Omnitracs’ Critical Event Reporting is one such solution. According to Kevin Haugh, Chief Strategy and Product Officer for Omnitracs, “These are sophisticated models that look at telematics safety events but also look at the driver’s schedule. How long do they drive? Longer hours mean more fatigue. But it is not just the hours, it is the time of day when those hours are logged.”
A company called SmartDrive Systems goes even further in their efforts to predict accidents. They are enriching the telematics data with video feeds from road facing and interior facing cameras. The company is combining asset sensor with driver sensor data to do better driver safety analytics and ultimately make better predictions.
For example, a telematics solution would capture a hard braking event and classify it as a negative behavior. But SmartDrive’s CEO Steve Mitgang points out. “That is not necessarily indicative of bad driving. What if a hard braking and swerving event occurred so the driver could avoid an accident with a teenage driver that swerved into his lane?”
SmartDrive Systems is a predictive analytics supplier whose solution is based on a public cloud architecture. In other words, all of their customers’ data is captured by the company; they have telematics and video on four billion miles driven, of which they scored almost 200 million events. SmartDrive’s customer agreements allow them to analyze all the customer data so they can continue to improve their algorithms. “We are constantly tuning and improving our algorithms,” Mr. Mitgang said.
The combination of telematics and video has allowed their scientists to better interpret the telematics data. By reading the telematics data, and seeing what happened, they were able to determine, for example, that turning more than 165 degrees within a certain turning radius and time window was a risky U-turn on a roadway. SmartDrive also detected additional patterns to avoid false trigger activations in large open areas such as parking lots and truck stops.
Is the video analysis real-time? “It is near real time. If our analysis of the data depicts severe risk – for example, a collision or near collision – a fleet analyst can be alerted in a couple of seconds, and can view the video right away.”
But the videos can also be used after the fact. If a triggering event is detected – incidents such as hard braking, swerves, and unsafe turning – fleet analysts can view the video to see if the driver was responding appropriately to a bad situation or if distracted driving or too closely following another vehicle was a contributor to the problem. This becomes a coachable event.
Mr. Mitgang emphasized coaching for correcting these problems. “Human behavior is not clean,” Mr. Mitgang said. “The worst drivers one month may perform well next month. Every day is different. Context matters, we don’t necessarily know what is going on in the driver’s life. For the most part these are really good, professional drivers.” And just like professional football players look at the film on how they performed the day after a game, “truck drivers are professionals too. Looking at how you performed to improve your performance is just part of being a professional.”
After the fact analysis can have another benefit. In a litigious society, the video can help to prove the truck driver was not at fault. Car drivers can’t see around big trucks. If the car driver claims they had an accident because a truck swerved, for example, the video may show the truck swerved to avoid something. “This can help to exonerate trucking firms from false claims.”
SmartDrive’s customers may also elect to share the company’s analytics with outside companies. ProSight Specialty Insurance, for example, is willing to offer lower rates to carriers when data can substantiate that a culture of safety exists.
There are other areas where these kinds of analytics can save money. The segment of drivers seen as most likely to be in a collision also consume over 7.5 percent more fuel according to their statistics. And because this is a private cloud solution, there is an opportunity for maintenance departments to see how their expenditures compare to other similar types of firms, and for operations departments to see how they compare to peers in areas like idling and fuel consumption.
While the trucking industry has the reputation of being somewhat slow to adopt new technologies, when it comes to Big Data and predictive analytics, the trucking industry is on the cutting edge.
Stewart Jackson says
I like this technology! It is fact that if one watches a subjects behaviour for long enough to build up a history one could predict the next actions by that subject given a similar situation.
For this reason the airline pilot who crashes a plane is subject to very close investigation. Gestalt psychology gives a base to this analytic method of human behaviour.
Yvette Lagrois says
This technology is a great asset to monitor existing driving behaviour.
The core of the problem exists with the entry level training.
Drivers need to be versed in the rules of engagement in the art of commercial driving before it can be monitored with less probability of an event. Otherwise the technology will watch bad behaviour that equates to going to the school of hard knocks before getting it right.