Traffic Smoothing Controllers for Autonomous Vehicles Using Deep Reinforcement Learning and Real-World Trajectory Data

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lichtlé, Nathan, Jang, Kathy, Shah, Adit, Vinitsky, Eugene, Lee, Jonathan W., Bayen, Alexandre M.
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911760173236224
author Lichtlé, Nathan
Jang, Kathy
Shah, Adit
Vinitsky, Eugene
Lee, Jonathan W.
Bayen, Alexandre M.
author_facet Lichtlé, Nathan
Jang, Kathy
Shah, Adit
Vinitsky, Eugene
Lee, Jonathan W.
Bayen, Alexandre M.
contents Designing traffic-smoothing cruise controllers that can be deployed onto autonomous vehicles is a key step towards improving traffic flow, reducing congestion, and enhancing fuel efficiency in mixed autonomy traffic. We bypass the common issue of having to carefully fine-tune a large traffic microsimulator by leveraging real-world trajectory data from the I-24 highway in Tennessee, replayed in a one-lane simulation. Using standard deep reinforcement learning methods, we train energy-reducing wave-smoothing policies. As an input to the agent, we observe the speed and distance of only the vehicle in front, which are local states readily available on most recent vehicles, as well as non-local observations about the downstream state of the traffic. We show that at a low 4% autonomous vehicle penetration rate, we achieve significant fuel savings of over 15% on trajectories exhibiting many stop-and-go waves. Finally, we analyze the smoothing effect of the controllers and demonstrate robustness to adding lane-changing into the simulation as well as the removal of downstream information.
format Preprint
id arxiv_https___arxiv_org_abs_2401_09666
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Traffic Smoothing Controllers for Autonomous Vehicles Using Deep Reinforcement Learning and Real-World Trajectory Data
Lichtlé, Nathan
Jang, Kathy
Shah, Adit
Vinitsky, Eugene
Lee, Jonathan W.
Bayen, Alexandre M.
Systems and Control
Artificial Intelligence
Multiagent Systems
Designing traffic-smoothing cruise controllers that can be deployed onto autonomous vehicles is a key step towards improving traffic flow, reducing congestion, and enhancing fuel efficiency in mixed autonomy traffic. We bypass the common issue of having to carefully fine-tune a large traffic microsimulator by leveraging real-world trajectory data from the I-24 highway in Tennessee, replayed in a one-lane simulation. Using standard deep reinforcement learning methods, we train energy-reducing wave-smoothing policies. As an input to the agent, we observe the speed and distance of only the vehicle in front, which are local states readily available on most recent vehicles, as well as non-local observations about the downstream state of the traffic. We show that at a low 4% autonomous vehicle penetration rate, we achieve significant fuel savings of over 15% on trajectories exhibiting many stop-and-go waves. Finally, we analyze the smoothing effect of the controllers and demonstrate robustness to adding lane-changing into the simulation as well as the removal of downstream information.
title Traffic Smoothing Controllers for Autonomous Vehicles Using Deep Reinforcement Learning and Real-World Trajectory Data
topic Systems and Control
Artificial Intelligence
Multiagent Systems
url https://arxiv.org/abs/2401.09666