<abstract xmlns="http://www.w3.org/1999/xhtml"><p>Smart Cities promise to their residents, quick journeys in a clean and sustainable environment. Despite, the benefits accrued by the introduction of traffic management solutions (e.g. improved travel times, maximisation of throughput, etc.), these solutions usually fall short on assessing the environmental impact around the implementation areas. However, environmental performance corresponds to a primary goal of contemporary mobility planning and therefore, solutions guaranteeing environmental sustainability are significant. This study presents an advanced Artificial Intelligence-based (AI) signal control framework, able to incorporate environmental considerations into the core of signal optimisation processes. More specifically, a highly flexible Reinforcement Learning (RL) algorithm has been developed towards the identification of efficient but-more importantly-environmentally friendly signal control strategies. The methodology is deployed on a large-scale micro-simulation environment able to realistically represent urban traffic conditions. Alternative signal control strategies are designed, applied, and evaluated against their achieved traffic efficiency and environmental footprint. Based on the results obtained from the application of the methodology on a core part of the road urban network of Nicosia, Cyprus the best strategy achieved a 4.8% increase of the network throughput, 17.7% decrease of the average queue length and a remarkable 34.2% decrease of delay while considerably reduced the CO emissions by 8.1%. The encouraging results showcase ability of RL-based traffic signal controlling to ensure improved air-quality conditions for the residents of dense urban areas.</p></abstract>

Smart Cities promise to their residents, quick journeys in a clean and sustainable environment. Despite, the benefits accrued by the introduction of traffic management solutions (e.g. improved travel times, maximisation of throughput, etc.), these solutions usually fall short on assessing the environmental impact around the implementation areas. However, environmental performance corresponds to a primary goal of contemporary mobility planning and therefore, solutions guaranteeing environmental sustainability are significant. This study presents an advanced Artificial Intelligence-based (AI) signal control framework, able to incorporate environmental considerations into the core of signal optimisation processes. More specifically, a highly flexible Reinforcement Learning (RL) algorithm has been developed towards the identification of efficient but-more importantly-environmentally friendly signal control strategies. The methodology is deployed on a large-scale micro-simulation environment able to realistically represent urban traffic conditions. Alternative signal control strategies are designed, applied, and evaluated against their achieved traffic efficiency and environmental footprint. Based on the results obtained from the application of the methodology on a core part of the road urban network of Nicosia, Cyprus the best strategy achieved a 4.8% increase of the network throughput, 17.7% decrease of the average queue length and a remarkable 34.2% decrease of delay while considerably reduced the CO emissions by 8.1%. The encouraging results showcase ability of RL-based traffic signal controlling to ensure improved air-quality conditions for the residents of dense urban areas.

Evaluation of Reinforcement Learning Traffic Signalling Strategies for Alternative Objectives: Implementation in the Network of Nicosia, Cyprus

Transport and Telecommunication Journal

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Smart Cities promise to their residents, quick journeys in a clean and sustainable environment. Despite, the benefits accrued by the introduction of traffic...

Evaluation of Reinforcement Learning Traffic Signalling Strategies for Alternative Objectives: Implementation in the Network of Nicosia, Cyprus

Data publikacji: 26 lis 2020

Zakres stron: 295 - 302

DOI: https://doi.org/10.2478/ttj-2020-0024

Słowa kluczowe
Reinforcement learning, Traffic signal control, Traffic management, Air quality, Large-scale micro-simulation

© 2020 Haris Ballis et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Evaluation of Reinforcement Learning Traffic Signalling Strategies for Alternative Objectives: Implementation in the Network of Nicosia, Cyprus

Data publikacji: 26 lis 2020

Zakres stron: 295 - 302

DOI: https://doi.org/10.2478/ttj-2020-0024

Słowa kluczoweReinforcement learning, Traffic signal control, Traffic management, Air quality, Large-scale micro-simulation

© 2020 Haris Ballis et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Słowa kluczowe
Reinforcement learning, Traffic signal control, Traffic management, Air quality, Large-scale micro-simulation