Section: New Results
Perception and autonomous navigation
Participants : Laurent Bouraoui, Olivier Garcia, François Charlot, Fawzi Nashashibi, Paulo Lopes Resende.
For an autonomous navigation system to be efficient, it is necessary to have three robust and relevant sub-systems: perception, planning and control. Very new results were achieved in perception this year with the development of a fast and reliable SLAMMOT system based on a single beam laser sensor. This system allows the environment scanning and modeling for the purpose of safe navigation on roads. The systems makes use of the built 3D model to generate a safe and optimized trajectory with respect to some criteria. This path planner is another aspect of command systems dedicated to the generation of correct trajectories for both autonomous mobile robots (Cycabs) or traditional ground vehicles. Two canonical demonstrations were designed and first results exhibited.
Autonomous platooning and crossroads monitoring
Participants : Laurent Bouraoui, Olivier Garcia, François Charlot.
La Rochelle's international workshop on “The automation for urban transport” gave us the opportunity to test in situ the latest innovative solutions for public transportation. Our system consists in ensuring autonomous safe driving on a known track, platooning and crossroads monitoring. In this experiment, several Cycabs equipped with GPS devices and communication media are able to share their respective global positions, detect other Cycabs, form virtual or immaterial platooning and perform an intelligent monitoring of the crossroads such that no accidents could occur between different Cycabs. This approach needs a refined localization thanks to map-matching techniques, a good perception and detection system and a well controlled synchronization with priority management.
A new co-pilot system for autonomous driving
Participants : Fawzi Nashashibi, Paulo Lopes Resende.
We currently work on a generic planner capable of finding secure trajectories for both robots and intelligent vehicles. The framework of the European project “HAVE-IT” pushed us to adapt our existing partial motion planner (PMP) in order to adapt to the road driving context. The co-pilot to be designed for HAVE-IT is based on a two-levels architecture: a “strategy level” which decides the type of maneuver to be executed, and a “motion planning level” dedicated to find a collision-free trajectory that is feasible for the vehicle with respect to its kinematics capacities and non-holonomic constraints.
