Dynamic Scene Understanding for Autonomous vehicles: Analysis, Prediction and Collision Risk Assessment (Invited talk) - Archive ouverte HAL Access content directly
Conference Papers Year :

Dynamic Scene Understanding for Autonomous vehicles: Analysis, Prediction and Collision Risk Assessment (Invited talk)

(1)
1

Abstract

Motion Autonomy and Safety issues in Autonomous Vehicles are strongly dependent upon the capabilities and performances of Embedded Perception and Situation Awareness systems. Recent benign and severe accidents (e.g. Tesla or Uber) have shown that the level of safety obtained using currently tested autonomous driving systems is still insufficient. This talk discusses these important perception and safety issues, and present how they can be addressed using a Bayesian Sensor Fusion approach. It will be shown how such a system (1) represent a dynamic scene using free, static, dynamic and unknown components, (2) predict upcoming changes in the scene, and (3) estimates short-term collision risks (about 3s ahead). Such an approach has been developed and patented by Inria and IRT Nanoelec (French Technological Research Institute), and it has recently been transferred to Toyota and to an industrial company operating in the field of new mobility products (company name is confidential).
Not file

Dates and versions

hal-01970436 , version 1 (07-01-2019)

Identifiers

  • HAL Id : hal-01970436 , version 1

Cite

Christian Laugier. Dynamic Scene Understanding for Autonomous vehicles: Analysis, Prediction and Collision Risk Assessment (Invited talk). SMIV 2018 - Smart Mobility and Intelligent Vehicles Conference, Vedecom, Nov 2018, Versailles, France. ⟨hal-01970436⟩
150 View
1 Download

Share

Gmail Facebook Twitter LinkedIn More