Vehicle Behavior Prediction Using Deep Learning
Title: Vehicle Behavior Prediction Using Deep Learning
SNIC Project: Berzelius-2022-191
Project Type: LiU Berzelius
Principal Investigator: Erik Frisk <erik.frisk@liu.se>
Affiliation: Linköpings universitet
Duration: 2022-09-21 – 2023-04-01
Classification: 20201
Homepage: https://liu.se/medarbetare/erifr93
Keywords:

Abstract

Accurately predicting the intention of surrounding vehicles is crucial for producing a safe and efficient motion plan of an autonomous vehicle. In complex traffic situations where multiple agents interact but share no direct communication, behavior prediction helps to enhance the autonomous vehicle's perception of the surrounding environment. The use of learning-based methods for the solution of these problems is a promising research topic. This project aims to investigate how deep networks may be used to produce predictions of the surrounding environment to support autonomous decision-making. The networks are trained and evaluated on vehicle trajectory datasets, recorded from various traffic scenarios.