​Everyday road maintenance often involves spraying bitumen and spreading gravel on cracked parts of the road to resurface them. Secmair asked the engineers at the CEA Tech Pays de la Loire regional branch to come up with a system to automatically detect cracks to be filled in and use AI to control how much bitumen and gravel is applied and where.

Traditionally, the repair vehicles that do this work are equipped with cameras that capture images of the road ahead. The operator then uses a joystick to control where the bitumen and gravel are applied. The information controls the equipment at the back of the vehicle, which either opens to release material or remains closed depending on the areas selected by the operator. These kinds of repairs are carried out at a speed of around 3.5 kph.

Researchers from the CEA Tech Pays de la Loire regional branch developed software to handle this tedious task. A neural network “learned” to observe the road surface. Initially, the images captured by the cameras were analyzed to identify cracks. If a given “crack density” threshold is exceeded, the software recommends the cracks be filled, and sends a map of the area to be repaired to the vehicle’s system. At the end of the process, the operator can view the software’s recommendations on a display and manually override them.

Not only does this make the operator’s job easier and safer, it also reduces the amount of material used, which results in cost savings, too. Initial tests carried out in 2020 were encouraging. And, if more advanced tests scheduled for next spring go well, the solution could be on the market in 2021.