Interview with Marc, a PhD in signal processing at VEDECOM, who dreams of helping seniors maintain their mobility even if they no longer have the capacity to drive. For him, the autonomous vehicle is not just a technical product, it’s a conviction that has informed his work for over 12 years. His line of attack? Helping vehicles stay on course by detecting lane markings!
Signal processing: from snowpacks to lane markings
I’ve been passionate about robotics since childhood and obtained my engineering degree at the University of Technology of Belfort-Montbéliard (UTBM), where I specialised in signal processing. That’s when I became interested in autonomous vehicles. My first professional experiences, however, were in other fields such as the simulation of theatres of operations for the military and modelling snowpacks in Switzerland. I then wanted to work on road safety and joined IFSTTAR, the French Institute of Science and Technology for Transport, Development and Networks, where I began my research on the detection of lane markings to guide vehicles. When VEDECOM, which was then only Mov’eo Tech, launched a call for submissions of thesis projects, I presented my subject: a mix between my work on lane marker detection at IFSTTAR and the multi-agent system inspired by my research on snowpacks. It was chosen and I was recruited by VEDECOM.
Vehicle relocalisation using lane marker edge detection
Today, my main mission is vehicle relocalisation in lanes by detecting a constellation of lane markers. In 2018, I developed an initial version of the software for this purpose. A second version is currently underway, which integrates 3D detection maps and new sources of information: a camera at the back of the vehicle is used with one placed at the front. This allows us to solve the problems caused by the sun at a low angle or glare in tunnels. This requires a lot of work integrating various sensors to ensure accurate calibration: time stamping and data recording, synchronising sensors and studying their operation so their use can be optimised.
I also give an “Introduction to the Autonomous Vehicle” training as part of the Institute’s training programme. I’m versatile in terms of autonomous vehicles and it seemed obvious to me to pass on my knowledge.
Adapting roads to autonomous vehicles: an essential precondition at no extra cost
One of the biggest challenges I have at the moment is being able to convince decision-makers that roads should be adapted to autonomous vehicles. It’s not asking much: only slight changes, at no additional cost. For example, changing the configuration of lane markings, which involves bringing marker edges closer together, changing their colour from white to green to avoid the marker artefacts that create shadows on the ground. This would already solve one of the two major problems associated with autonomous vehicles: centimetric localisation of vehicles. Technically speaking, we will still need to overcome the difficulties of detection caused by shadows on the ground and heavy rain, not to mention the snow that’s likely to make us switch back to manual mode.
Building a strong team at VEDECOM around the autonomous vehicle
VEDECOM has a unique position as a knowledge-sharing centre for all French players in the autonomous vehicle sector. That’s very motivating, as is the great flexibility it offers researchers. We feel invested in what we do. I’m one of the first people who joined VEDECOM and my reason for being here has always been to try to create a strong team around the autonomous vehicle. It’s a field where all the skills are interdependent and the notion of team is essential. Basically, when obstacle detection doesn’t work, it has a negative effect on marker detection.
Exciting projects with loads of potential
The contactless dynamic charging demonstrations of the European FABRIC project, conducted with all our partners, were clearly very motivating. Our system was used concretely to help the vehicle localise itself as precisely as possible on the induction charging coil integrated into the road in order to optimise coupling with the vehicle’s coil as it moves forward. And it worked! We’re going to continue working on this project within the framework of the European INCIT-EV project and this time we’ll even try and control the steering wheel to automate vehicle guidance!
Another highlight for me happened at the end of a presentation to partners, one of whom told me: “At first, I didn’t believe what you explained to me. You kept going despite everything and finally today you’ve shown us something that has us convinced. We’ll continue next year, as we’re making a V2 and we’d like you to come and integrate it into our vehicles.” This kind of recognition from manufacturers is incredibly encouraging!