LibSignal: An Open Library for Traffic Signal Control

TL;DR

LibSignal is an open-source library that enables fair comparison of reinforcement learning models for traffic signal control across multiple simulators and datasets, facilitating standardized evaluation.

Contribution

It provides a unified platform with extensible interfaces and evaluation metrics for cross-simulator comparison of traffic signal RL models, which was not available before.

Findings

Validated RL models across SUMO and CityFlow

Calibrated simulators for comparable results

Compared state-of-the-art RL algorithms fairly

Abstract

This paper introduces a library for cross-simulator comparison of reinforcement learning models in traffic signal control tasks. This library is developed to implement recent state-of-the-art reinforcement learning models with extensible interfaces and unified cross-simulator evaluation metrics. It supports commonly-used simulators in traffic signal control tasks, including Simulation of Urban MObility(SUMO) and CityFlow, and multiple benchmark datasets for fair comparisons. We conducted experiments to validate our implementation of the models and to calibrate the simulators so that the experiments from one simulator could be referential to the other. Based on the validated models and calibrated environments, this paper compares and reports the performance of current state-of-the-art RL algorithms across different datasets and simulators. This is the first time that these methods have been compared fairly under the same datasets with different simulators.