From a video camera to a driverless shuttle

A new type of driverless shuttle has been designed thanks to innovative computer vision guidance technology which is an inexpensive and reliable system that enables the vehicle to locate itself on the road. This technology is based on the use of simple video cameras and has been developed by Institut Pascal researchers as part of the LabEx IMobS3 project. It is at the heart of the EZ-10 autonomous shuttle developed by Ligier Group, unveiled at the recent Michelin Challenge Bibendum in Chengdu (China).

Institut Pascal / LabEx IMobS3 (In french)

For almost two decades, many companies have been working towards creating autonomous vehicles using expensive and sometimes unreliable technologies1. In 2003, researchers from Institut Pascal decided to start working on the automated driving of electric urban vehicles using simple video cameras. The technology they have developed is based on two stages. The first aims to make note of any remarkable aspects of the environment surrounding the planned route, based on video footage filmed during an initial manual drive-through. The second is the automatic mode, when the vehicle continuously controls its trajectory in such a way as to ensure the images provided by the onboard cameras match as closely as possible with the initial footage. In this way, the initial video serves as a virtual railway that the vehicle must follow during its autonomous drive mode.

Since 2006, researchers from Institut Pascal have been collaborating with Ligier Group to design automatic driverless shuttles that can carry up to 10 people, much like a horizontal elevator for short trips (less than a mile). These vehicles are designed to be deployed in dedicated areas (industrial sites, airports, theme parks, etc.) and can detect potential dangers thanks to laser rangefinders located on all four corners of the vehicle. The shuttle can also detect obstacles 262 feet away and adjust its speed or even stop, depending on the degree of risk. Researchers from LabEx IMobS3 are also studying the optimal management of a fleet of vehicles. Their goal: to deal with multiple, potentially simultaneous requests coming from call boxes or smartphones, in real-time and across a large area, much like a fleet of automatic taxis.

1 Example: differential GPS systems (which differ from commonly-used car GPS systems) are not always reliable, especially in city centers, where satellite signals bounce of building facades. This phenomenon leads to errors in location calculations.