Features
7 Jan 19

Artificial intelligence and Fleets

At a conference on Artificial Intelligence held by fleet software creator GAC Technology in Lyon at the end of the year, the impact of AI on fleet management was investigated. 

Firstly, and for those among us who may not appreciate the difference between AI and the sort of work our computers have always done, there was some explanation. AI involves creating devices which can do things that, if they were to be done by humans, would require intelligence. The original input, however, still comes from humans – you have to teach the device to do what you want it to do. This is achieved by programming the device to deal with gigantic amounts of data in a particular way, taking parameters into account which an intelligent human would do automatically.

The speakers at the conference had a word of comfort, pointing out that AI will never replace humans, just cause their roles to change. The analogy used was that of the fact that accountants no longer have to add up figures themselves, machines have been taught to do this. There are also limits to what a machine can do, compared to a human. The example given was that if you ask a robot telephone receptionist what time it is, it will say ‘Eleven o’clock’. And if you ask it what the weather is going to be, it will say ‘It’s going to rain’. But if you ask it what time it is and what the weather is going to be, it can’t answer.

Intelligent analysis

Moving on to the impact of AI on transport and fleets, a first example is merchandise. Taking all types of parameters into account, an AI device will be able to select the best transport company for the job, and thus bring down costs. For car fleets and insurance, an AI device can handle a huge amount of accident data in one day, which humans can’t, again saving money and time. If correctly programmed, an AI device can also begin to detect trends, as humans do, and therefore even detect fraud and suggest that a human operator examines the case in question.

Combined data

A speaker from tyre manufacturer Michelin said that data collected from cars via their in-built systems, along with data from other sources such as GPS, can be combined and the whole data set thus enriched. This other data may include, for example, the time of day, the type of road, the weather – which non-AI devices do not take into account. With all this information available, the device can then distinguish between two apparently identical events. A driver who brakes before a bend has anticipated a situation while one who brakes in the bend hasn’t… This will lead to fleet managers receiving different types of advice for drivers experiencing the same event, depending on the usage profile. 

Intelligent analysis

In terms of concrete fleet management, GAC Technology Director General Matthieu Echalier explained how the latest systems being developed by his company will contribute. For example, in terms of making sense of mileage recordings which are clearly out of line with the norm, an AI device can automatically do the smoothing, and can even recognise future inconsistencies, removing them from the equation before providing useable data to the fleet manager.

In a similar vein, the system can use its AI to analyse the reasons behind abnormally high or low fuel spends within a fleet. It will, for example, know that there are more electric cars in the fleet, bringing fuel costs down. It will also know that there has been a change in the way the fleet is handled financially, affecting the way the figures are reported. And an AI device will know that in August, many drivers are on holiday, so it ‘won’t wake the fleet manager up in the middle of the night to tell him that the fuel spend is abnormally low…’

Matthieu Echalier also said that where drive behaviour is concerned, the AI system will take different types of parameters into account, differentiating between a parking offence, speeding… It can then analyse thousands of pieces of data from insurance companies among others, and, along with the categorisation of offence data, produce risk analyses and risk curves. Alert thresholds for fleet managers are therefore personalised by driver.

Image: the event was held in the prestigious surroundings of the Lyon Opéra
 

Authored by: Tim Harrup