STM Journals Peer-reviewed research from Consortium e-Learning Network

Journal of Communication Engineering & Systems

A Machine Vision Approach for Geofence-based Adaptive Friction Modulation in Connected Semi-autonomous Vehicles

  1. Ushaa Eswaran
  2. Vivek Eswaran
  3. Keerthna Murali
  4. Vishal Eswaran

Abstract

Safeguarding vulnerable road user (VRU) safety necessitates fail-safes preventing collisions from semi-autonomous vehicles. Geofencing allows service-zones definition but introduces friction discouraging adoption. This paper details an onboard vision model detecting VRUs by helmet status, triggering reactionary neuro-modulations of vehicle maneuverability. The architecture encompasses a YOLOv3 detector determining pedestrian/bicyclist crossing intent through path extrapolations, calibrated using lidar-fusion correcting distance estimates. The subsequent velocity governor adapts acceleration profiles based on predicted exposure risks, verified through hardware-in-loop testing. Ease-of-use implementations will encourage voluntary adoption of potentially life-saving technologies meeting pedestrian fatality reduction goals.

Keywords

Support