Citation

Englund C. Int. J. Veh. Des. 2020; 83(2/3/4): 103-121.

Copyright

(Copyright © 2020, Inderscience Publishers)

DOI

10.1504/IJVD.2020.115056

PMID

unavailable

Abstract

Action intention recognition is becoming increasingly important in the road vehicle automation domain. Autonomous vehicles must be aware of their surroundings if we are to build safe and efficient transport systems. This paper presents a method for predicting the action intentions of road users based on sensors in the road infrastructure. The scenarios used for demonstration are recorded on two different public road sections. The first scenario includes bicyclists and the second includes cars that are driving in a road approaching an intersection where they are either leaving or continuing straight. A 3D camera-based data acquisition system is used to collect trajectories of the road users that are used as input for models trained to predict the action intention of the road users. The proposed system enables future connected and automated vehicles to receive collision warnings from an infrastructure-based sensor system well in advance to enable better planning.

Keywords: intention recognition; random forest; data mining; traffic behaviour modeling; variable selection.

Language: en