Deep RL reward function design for lane-free autonomous driving

This item is provided by the institution :
Technical University of Crete

Repository :
Institutional Repository Technical University of Crete

see the original item page
in the repository's web site and access all digital files if the item^*

Title

Deep RL reward function design for lane-free autonomous driving (EN)

Creator

Τρουλλινος Δημητριος (EL)

Χαλκιαδακης Γεωργιος (EL)

Παπαγεωργιου Μαρκος (EL)

Καραλακου Αθανασια (EL)

Chalkiadakis Georgios (EN)

Karalakou Athanasia (EN)

Papageorgiou Markos (EN)

Troullinos Dimitrios (EN)

Type

full paper

conferenceItem

Conference item (EN)

Article
Scientific article (EN)

Issued

2022

Year

2022 (EN)

Description

In this paper we present an application of Deep Reinforcement Learning to lane-free traffic, where vehicles do not adhere to the notion of lanes, but are rather able to be located at any lateral position within the road boundaries. This constitutes an entirely different problem domain for autonomous driving compared to lane-based traffic, as vehicles consider the actual two dimensional space available, and their decision making needs to adapt to this concept. We also consider that each vehicle wishes to maintain a (different) desired speed, therefore creating many situations where vehicles need to perform overtaking, and react appropriately to the behaviour of others. As such, in this work, we design a Reinforcement Learning agent for the problem at hand, considering different components of reward functions tied to the environment at various levels of information. Finally, we examine the effectiveness of our approach using the Deep Deterministic Policy Gradient algorithm. (EN)

Subject

Deep reinforcement learning (EN)

Lane-free traffic (EN)

Autonomous driving (EN)

Conference

20th International Conference on Practical Applications of Agents and Multi-Agent Systems (EL)

Language

English

Publisher

Springer (EN)

School / Department / Institute

Πολυτεχνείο Κρήτης (EL)