Startup Studio
A driving coach for automated vehicles
Date:
Changed on 09/04/2025
A new company project at Inria's Startup Studio, Mobpti is developing a software solution to increase confidence in the decision-making algorithms of autonomous vehicles. The result: an easier validation process for automakers.
Alexis is an engineer with a PhD in robotics. He completed his thesis at LAAS-CNRS/INP Toulouse on estimation and stabilization for humanoid robots. He then spent three years working as a research engineer at the Inria center at the University of Grenoble-Alpes. There, he developed trajectory generation algorithms for autonomous buses, using MPC (Model Predictive Control) techniques. He then worked as a post-doctoral fellow at ENSTA Paris and as a research engineer at LS2N, École Centrale de Nantes. Today, he is returning to his work at the Inria center at the University of Grenoble-Alpes to develop this project to help validate automated vehicles.
"Manufacturers planning to bring self-driving vehicles to market must undergo extensive verification, validation, homologation and certification phases. This is both time-consuming and costly. The software we are developing will enable us to anticipate and therefore reduce potential failures during these stages," sums up Alexis Mifsud. The research engineer behind the Mobpti start-up project has just joined the Startup Studio at the Inria Center at Rennes University for a year-long technological maturation cycle. Here, he is fine-tuning the algorithms he developed during his time at Grenoble University's Inria Center, in a team working on trajectory generation for autonomous buses.
The tool differs from embedded software in two respects.
I don't have to find a solution very quickly. Which, of course, is what automakers have to do with their vehicles. I also remove the constraint of limited computing power. I can afford to spend two hours calculating a solution for a scenario, where normally you need to find an answer within a millisecond.
Identify the best driving solutions
Thanks to all this, Mobpti will be able to identify the best driving solutions for given scenarios. “For example, depending on the environment, we can say that at 26 km/h, the vehicle has sufficient braking margin if an obstacle is encountered at a given distance. This would no longer be the case at 32 km/h. This is the type of information we can produce. We can do this by taking into account the road profile, several possible futures for dynamic obstacles...”
The proposed service will consist of two software packages. The first is accessible online. “It looks for optimal solutions, the best response to a given driving situation. It deduces a measure of the unavoidable risk associated with this scenario.” The second operates locally on the user's machine. “It too performs the measurement, but this time with the manufacturer's solution. We then subtract between the two.”
Image
Verbatim
The resulting information will enable development teams to identify areas for improvement in the security and robustness of their application. It's worth noting that the entire process is carried out without having to access the manufacturer's source code. No risk to the manufacturer's intellectual property.
Auteur
Poste
Mobpti project leader at Inria Startup Studio and research engineer
Towards a brick that can be integrated into simulators
Who will this new tool be aimed at ? Autonomous vehicle manufacturers ? “Yes, but also test centers. These are key players in the industry. The State delegates to them the responsibility of organizing the homologation task, for example. To do their job, they themselves use simulation software such as Carla, IPG Automotive or AV Simulation. Eventually, our brick could even be integrated into some of these simulators, as we complement each other. The aim of these programs is to get as close as possible to reality. They even simulate the stiffness of suspension springs. Conversely, our tool doesn't need this level of detail. By focusing only on the decision-making part of vehicles, which is increasingly important and intelligent, we can simplify the problem and provide this complementarity.”
Public transport
The project aims to target public transport. “It's the mode of transport most in line with our ethical values and the general interest. If, in our cities, everyone moves around using a personal vehicle with totally autonomous driving, we're creating an ecological, social and economic aberration.” That said, “our solution is also of interest to carmakers, as the safety of their driving aids is a central concern for them.”
The maturation phase has just begun at Inria's Startup Studio, and is accelerating with the first recruitment of an engineer. “It should enable us to present a proof of concept in about a year's time. Until then, we plan to take part in the UTAC Challenge in May!”