An Open-Source Framework for Adaptive Traffic Signal Control (1909.00395v1)

Abstract: Sub-optimal control policies in transportation systems negatively impact mobility, the environment and human health. Developing optimal transportation control systems at the appropriate scale can be difficult as cities' transportation systems can be large, complex and stochastic. Intersection traffic signal controllers are an important element of modern transportation infrastructure where sub-optimal control policies can incur high costs to many users. Many adaptive traffic signal controllers have been proposed by the community but research is lacking regarding their relative performance difference - which adaptive traffic signal controller is best remains an open question. This research contributes a framework for developing and evaluating different adaptive traffic signal controller models in simulation - both learning and non-learning - and demonstrates its capabilities. The framework is used to first, investigate the performance variance of the modelled adaptive traffic signal controllers with respect to their hyperparameters and second, analyze the performance differences between controllers with optimal hyperparameters. The proposed framework contains implementations of some of the most popular adaptive traffic signal controllers from the literature; Webster's, Max-pressure and Self-Organizing Traffic Lights, along with deep Q-network and deep deterministic policy gradient reinforcement learning controllers. This framework will aid researchers by accelerating their work from a common starting point, allowing them to generate results faster with less effort. All framework source code is available at https://github.com/docwza/sumolights.

• The paper presents a unified open-source framework that evaluates both heuristic and reinforcement learning controllers for adaptive traffic signal control.
• It employs SUMO microsimulation and parallel computing to optimize hyperparameters and analyze performance in dynamic traffic scenarios.
• Experimental results reveal Max-pressure's efficiency and highlight the variability of learning-based controllers, emphasizing areas for further improvement.

An Open-Source Framework for Adaptive Traffic Signal Control: A Comprehensive Analysis

The paper "An Open-Source Framework for Adaptive Traffic Signal Control" by Wade Genders and Saiedeh Razavi delineates the development and evaluation of an open-source framework designed to enhance the paper and implementation of adaptive traffic signal control systems. Traffic signals are pivotal elements in urban transportation infrastructure, and sub-optimal control can exacerbate fuel consumption, emissions, and congestion. This research aims to accelerate the development and assessment of adaptive traffic signal controllers by providing a cohesive environment for testing different models, both learning-based and heuristic.

Framework and Methodologies

The framework consolidates several well-recognized adaptive traffic signal controllers, including Webster's method, Max-pressure, Self-Organizing Traffic Lights (SOTL), and reinforcement learning approaches utilizing Deep Q-Networks (DQN) and Deep Deterministic Policy Gradient (DDPG) methods. It is optimized for the SUMO traffic microsimulator and employs parallel computing to handle complex networks efficiently.

The framework is bifurcated into non-learning and learning categories. Non-learning controllers, such as Webster's and Max-pressure, employ fixed rules or heuristics to determine traffic signal phases, allowing for scalable and computationally efficient control policies. On the other hand, learning-based controllers like DQN and DDPG utilize reinforcement learning to adapt traffic signals based on real-time conditions, leveraging neural networks to model policy and value functions.

Experimental Results

A salient aspect of the paper is the examination of hyperparameter sensitivity across different adaptive signal control models. The paper employs a grid search methodology to optimize these hyperparameters and assess their impact on traffic efficiency metrics such as travel time, queue length, and delay across a two-intersection scenario. The results underscore the importance of hyperparameter optimization, particularly for learning-based controllers, which show greater sensitivity to parameter tuning compared to heuristic-based systems.

The Max-pressure controller consistently demonstrates superior performance with low travel times and variability, marking its efficacy in handling dynamic traffic conditions. In contrast, the reinforcement learning controllers (DQN and DDPG) exhibit greater performance variation, indicating potential areas for further exploration and refinement.

Theoretical and Practical Implications

The research has significant implications for transportation systems engineering. Practically, it enables rapid prototyping and evaluation of adaptive signal control algorithms in simulated environments, reducing the resource investment required for researchers and practitioners. Theoretically, it contributes to the understanding of how adaptive systems can be dynamically configured to respond to real-world challenges, laying the groundwork for future innovations in intelligent transportation systems.

Future developments could focus on integrating richer environmental representations and enhancing the scalability of learning algorithms to accommodate larger networks. The exploration of novel reinforcement learning techniques—such as distributional reinforcement learning—could potentially yield models capable of outperforming current heuristics like Max-pressure.

In summary, the paper provides an expansive framework facilitating the evaluation and deployment of adaptive traffic signal controllers, offering a valuable asset to researchers seeking to optimize urban traffic management. The insights from the evaluation of learning versus non-learning controllers reveal a promising avenue for future research, emphasizing the ongoing evolution of adaptive control in complex, stochastic environments of urban transportation systems.