Some telematics suppliers now offer artificial intelligence (AI)-driven data analysis, which claims to make things much quicker and simpler for all involved, by narrowing the collected information down to what is actually relevant. Let’s imagine a hypothetical scenario where a fleet vehicle is discovered with unexplained body damage. For this case, two suppliers discuss what modern products can do to help pin down what really happened – and perhaps how to avoid it happening in future.
Damian Penney, VP EMEA at Lytx, says: ‘‘The integration of machine vision and artificial intelligence (MV+AI) technologies with video telematics platforms is removing the need for operators to review hours of footage. [It does this] by pinpointing risky driving behaviours that can then be flagged via a 12-second video clip. As well as saving valuable time, these technologies are also helping to prevent incidents from happening in the first place.”
The machine vision part of Lytx’s camera-based product is used to spot instances of behaviour that has been defined as risky – texting at the wheel, smoking, failing to stop at junctions or not wearing a seatbelt, for example. This means no human needs to physically watch all the footage in order to find things which might cause a problem. AI, meanwhile, determines the level of risk associated with these actions, which all have high correlation with incidents such as collisions occurring.
“When risk is detected, [the system] triggers an in-cab alert that empowers drivers to self-correct in the moment,” Penney says. “This helps to prevent a moment of distraction from leading to a serious incident.” If ‘risky’ behaviour is picked up repeatedly, alerts can be sent to the operator, along with short video clips to review. A decision can then be made on whether additional coaching or training is required.
“Upon reviewing in-cab and auxiliary footage, the operator might see that another driver was swerving or driving too close and bumped into the vehicle in front, causing the dent,” he says, referring to the hypothetical scenario. ‘‘When an incident occurs, video tells us how and why – which is important in providing an accurate view of what’s really happened and exonerating any drivers who may have been wrongfully blamed.”
What Penney doesn’t make clear, however, is how to find that incident in the first place, meaning fleet managers must rely on it having been serious or dramatic enough for the machine vision to have picked up. This is fine in the case of a significant collision or when contact has been made fairly violently with a solid object, but unlikely to help in cases where corner panels have been scuffed on gateposts or inadvertently grazed along a parked car while slow speed manoeuvring – exactly the kind of thing a driver may decide to keep quiet about.
PUTTING ON THE BRAKES
Truck rental company Fraikin offers its Mysmartfleet digital fleet management system to all customers, within which is a platform called Smartvision. “We use telematics that connect directly to the braking system on the vehicle,” says technology solutions manager Regan Greeff, “and to that add either road facing-only or dual-facing cameras.” There is also the option of integrating third-party cameras to the system for extra detection, allowing the use of a single sign-on to multiple platforms.
Like other AI-based systems, Fraikin’s product uses algorithms to detect certain events that can be followed up, but it then goes on to work with drivers to stop mistakes being repeated. As well as alerting operators, drivers can be virtually coached, with repeat offenders also being listened to and their style of driving evaluated. “Of course, driving safety is paramount, as is mental wellbeing,” Greeff reminds us. “So it is key to understand the circumstances surrounding any incident.” There are GDPR implications to the collection and sharing of data like this, too, so face blurring for drivers, passengers and pedestrians is available as required.
Having identified an incident, the first option is to have the driver themselves view the footage, to see what may have gone wrong. “The other bit is looking at a pattern of behaviour,” Greeff says. “If one person has these events on a routine basis, perhaps a driver-trainer should be speaking to them and encouraging them to see for themselves if there was anything they could have done differently.”
The use of AI to direct all of this reduces the amount of footage and data the manager or operator is faced with, while engagement with drivers works towards a solution. “Because ultimately change comes from within,” Greeff adds. “It may be mandated from the top, but if you don’t have engagement from the bottom, nothing will go anywhere.”
Returning to our mystery dent, especially the kind which may not trigger a harsh braking, acceleration or steering event, Greeff has a different suggestion. “Firstly, blame culture is not desirable in any organisation, and I would encourage anyone to stay away from that mindset,” he tells us. “Secondly, the most important way to mitigate things like that is going to be your driver walk-around checks.” Doing these digitally means a record of vehicle condition can be made before and after each shift, with the option to take pictures.
This would seem to suggest that there is no one magic bullet to solve every operator problem of this nature. “I would say that having more items at your disposal certainly makes it easier, but customers buy these systems and think that just because they’ve bought it, it’s on the vehicle and everything’s going to go swimmingly,” says Greeff, who has also worked for well-known supplier MIX Telematics in the past. “What they fail to realise is that engagement, both from an operating perspective as well as on the driver side, together creates the solution.”
BOX: WHO IS RESPONSIBLE?
Last year, criminal and regulatory lawyer and barrister Charlotte Le Maire, part of the LMP Legal Partnership, discussed how having telematics and cameras fitted to vehicles can actually work against operators in court if the data from these is not properly monitored (see also
www.is.gd/abudud). Will an AI-based system tick that box? “If an operator outsources, this is a useful way of ensuring that the data gathered is being looked at without utilising their own resource,” Le Maire states. “However, that isn’t the end of their responsibilities.“ These systems will still generate trends and data, which ultimately the operator must then look at and take action on. “They will still need to engage with the reports and be the one to act on any issues uncovered,” she concludes. “There really isn’t a way around an operator taking time to consider the data, in one form or another.”