Highway behaviour training through learning based state choice model

János Szalay, Bálint Kővári, Szilárd Aradi, Péter Gáspár, Tamás Bécsi

Research output: Contribution to conferencePaper

Abstract

Automated highway driving solutions are reaching the commercially available vehicles, though with restricted functionalities. The research presents a mixed simulation model, where the behavior planning of a vehicle can be designed with interactive traffic environment, the traffic is modelled with classic microscopic approach, though the vehicle to be controlled uses more complex model. The control of the vehicle is based on a hierarchical state machine approach, where the model can choose from different basic behavior schemes, like lane keeping, lane changing etc. with underlying low-level control for all states. The paper shows the performance of the classic state choice models and compares it to a machine learning based approach, where the state transition decisions are controlled by a neural network structure.

Original languageEnglish
Pages417-423
Number of pages7
Publication statusPublished - Jan 1 2019
Event16th Mini Conference on Vehicle System Dynamics, Identification and Anomalies, VSDIA 2018 - , Hungary
Duration: Nov 5 2018Nov 7 2018

Conference

Conference16th Mini Conference on Vehicle System Dynamics, Identification and Anomalies, VSDIA 2018
CountryHungary
Period11/5/1811/7/18

    Fingerprint

Keywords

  • Highway behaviour
  • Machine learning

ASJC Scopus subject areas

  • Mechanical Engineering

Cite this

Szalay, J., Kővári, B., Aradi, S., Gáspár, P., & Bécsi, T. (2019). Highway behaviour training through learning based state choice model. 417-423. Paper presented at 16th Mini Conference on Vehicle System Dynamics, Identification and Anomalies, VSDIA 2018, Hungary.