Embedded Bayesian Perception for Autonomous Driving - Archive ouverte HAL Access content directly
Conference Papers Year : 2017

Embedded Bayesian Perception for Autonomous Driving



The main issues addressed in this talk are the followings: •Embedded Perception is one of the major bottlenecks for Motion Autonomy, and Safety is still insufficient. New solutions have still to be investigated. •The current lack of Robustness, Efficiency and Software / Hardware integration of these systems is still an obstacle to a full deployment of autonomous driving technologies. These important issues have to be more deeply addressed. •An novel approach, called "Bayesian Perception", is proposed to improve the system Robustness and Efficiency in highly dynamic and uncertain environments. •Another important property of this approach is to include a "prediction step", which allows the system to continuously evaluate the "Risk of Collision" with the surrounding traffic participants (for safe navigation). •The Real-time implementation issue will also be discussed.
Not file

Dates and versions

hal-01672171 , version 1 (23-12-2017)


  • HAL Id : hal-01672171 , version 1


Christian Laugier. Embedded Bayesian Perception for Autonomous Driving. IS Auto Europe 2017, Apr 2017, Dusseldorf, Germany. ⟨hal-01672171⟩
189 View
0 Download


Gmail Facebook Twitter LinkedIn More