Hybrid FPA: a fairness-, pressure-, and emission-aware deep reinforcement learning framework for adaptive traffic signal control
Mihir R. Panchal1 and Pankaj P. Prajapati2
Associate Professor, Department of Electronics and Communication Engineering,Vishwakarma Government Engineering College, Gujarat Technological University, Ahmedabad,Gujarat,India2
Corresponding Author : Mihir R. Panchal
Recieved : 22-August-2025; Revised : 19-April-2026; Accepted : 20-April-2026
Abstract
In urban networks, signalized intersections represent critical bottlenecks that often lead to fuel wastage, prolonged delays, and increased emissions of carbon dioxide (CO₂), carbon monoxide (CO), nitrogen oxides (NOx), hydrocarbons (HC), and particulate matter (PMx). Conventional fixed-time and actuated controllers tend to degrade under nonstationary demand, heterogeneous traffic composition, and recurrent traffic surges, resulting in congestion and environmental degradation. This study proposes a hybrid flow pressure-aware (Hybrid FPA) framework, a fairness–pressure–aware deep reinforcement learning (DRL) approach that integrates switching penalties, emission-weighted pressure, and fairness regularization into a unified reward function. The framework is supported by a double dueling deep Q-network (D3QN) with prioritized experience replay (PER) for enhanced stability, along with a compact one-dimensional convolutional (Conv1D) encoder. Using traffic data collected from unmanned aerial vehicles (UAVs), the proposed framework is evaluated on a Simulation of Urban MObility (SUMO)-based reconstruction of the Panjarapol Cross Road intersection in Ahmedabad. The performance is compared with state-of-the-art methods, including PressLight, convolutional block attention module D3QN (CBAM-D3QN), multi-directional D3QN (MD3QN), and partial detection D3QN (PD3QN). Experimental results over 100 training episodes demonstrate that Hybrid FPA reduces average waiting time by approximately 78% and queue length by about 65%, while also decreasing fuel consumption by nearly 27% and CO₂ emissions by 29%. Additionally, PMx emissions are reduced by nearly 20%, HC by over 20%, and NOx by approximately 25%, along with reductions in other pollutants. The phase-switching frequency is also lowered by more than one-third, contributing to smoother traffic flow and further indirect emission reductions. These findings indicate that reinforcement learning can effectively achieve both mobility and sustainability objectives by explicitly incorporating fairness, pressure balancing, and multi-pollutant optimization. Therefore, Hybrid FPA represents a dynamic and efficient paradigm for adaptive traffic signal control with strong potential for real-world implementation.
Keywords
Hybrid flow pressure-aware (Hybrid FPA), Deep reinforcement learning (DRL), Traffic signal control, Emission reduction, Intelligent transportation systems, SUMO simulation.
Cite this article
Panchal MR, Prajapati PP. Hybrid FPA: a fairness-, pressure-, and emission-aware deep reinforcement learning framework for adaptive traffic signal control. International Journal of Advanced Technology and Engineering Exploration. 2026;13(137):541-565. DOI : 10.19101/IJATEE.2025.121221179
[1] Van DBP, Arentze T, Timmermans H. A path analysis of social networks, telecommunication and social activity–travel patterns. Transportation Research Part C: Emerging Technologies. 2013; 26:256-68.
[2] Mnih V, Kavukcuoglu K, Silver D, Rusu AA, Veness J, Bellemare MG, et al. Human-level control through deep reinforcement learning. Nature. 2015; 518(7540):529-33.
[3] Van HH, Guez A, Silver D. Deep reinforcement learning with double q-learning. In proceedings of the AAAI conference on artificial intelligence 2016 (pp. 2094-100). PKP.
[4] Wang Z, Schaul T, Hessel M, Hasselt H, Lanctot M, Freitas N. Dueling network architectures for deep reinforcement learning. In international conference on machine learning 2016 (pp. 1995-2003). PMLR.
[5] Horgan D, Quan J, Budden D, Barth-maron G, Hessel M, Van HH, et al. Distributed prioritized experience replay. In international conference on learning representations 2018 (pp. 1-19).
[6] Wei H, Zheng G, Yao H, Li Z. Intellilight: a reinforcement learning approach for intelligent traffic light control. In proceedings of the 24th SIGKDD international conference on knowledge discovery & data mining 2018 (pp. 2496-505). ACM.
[7] Wei H, Chen C, Zheng G, Wu K, Gayah V, Xu K, et al. Presslight: learning max pressure control to coordinate traffic signals in arterial network. In proceedings of the 25th SIGKDD international conference on knowledge discovery & data mining 2019 (pp. 1290-8). ACM.
[8] Zhao H, Dong C, Cao J, Chen Q. A survey on deep reinforcement learning approaches for traffic signal control. Engineering Applications of Artificial Intelligence. 2024; 133:108100.
[9] Zhang K, Cui Z, Ma W. A survey on reinforcement learning-based control for signalized intersections with connected automated vehicles. Transport Reviews. 2024; 44(6):1187-208.
[10] Michailidis P, Michailidis I, Lazaridis CR, Kosmatopoulos E. Traffic signal control via reinforcement learning: a review on applications and innovations. Infrastructures. 2025; 10(5):1-41.
[11] Zhang G, Chang F, Huang H, Zhou Z. Dual-objective reinforcement learning-based adaptive traffic signal control for decarbonization and efficiency optimization. Mathematics. 2024; 12(13):2056.
[12] Haydari A, Aggarwal V, Zhang M, Chuah CN. Constrained reinforcement learning for fair and environmentally efficient traffic signal controllers. Journal on Autonomous Transportation Systems. 2024; 2(1):1-9.
[13] Wang P, Ni W. An enhanced dueling double deep Q-network with convolutional block attention module for traffic signal optimization in deep reinforcement learning. IEEE Access. 2024; 12:44224-32.
[14] Ducrocq R, Farhi N. Deep reinforcement Q-learning for intelligent traffic signal control with partial detection. International Journal of Intelligent Transportation Systems Research. 2023; 21(1):192-206.
[15] Fan S, Lu K, Wang Y, Tian X, Zhang M. Action masking-based proximal policy optimization with the dual-ring phase structure for adaptive traffic signal control. IEEE Transactions on Intelligent Transportation Systems. 2024; 26(2):2422-33.
[16] Obaid L, Hamad K, Al-ruzouq R, Dabous SA, Ismail K, Alotaibi E. State-of-the-art review of unmanned aerial vehicles (UAVs) and artificial intelligence (AI) for traffic and safety analyses: recent progress, applications, challenges, and opportunities. Transportation Research Interdisciplinary Perspectives. 2025; 33:101591.
[17] Sarkar DR, Rao KR, Chatterjee N. Automatic traffic safety analysis using unmanned aerial vehicle technology at unsignalized intersections in heterogeneous traffic. Transportation Research Record. 2025; 2679(2):1274-90.
[18] Deng Y, Garoni TM, Grimm J, Nasrawi A, Zhou Z. The length of self-avoiding walks on the complete graph. Journal of Statistical Mechanics: Theory and Experiment. 2019; 2019(10):103206.
[19] Guan J, Yang X, Liu P, Oeser M, Hong H, Li Y, et al. Multi-scale asphalt pavement deformation detection and measurement based on machine learning of full field-of-view digital surface data. Transportation Research Part C: Emerging Technologies. 2023; 152:104177.
[20] Li C, Yan H, Zhao Q. Efficient policy transfer in large-scale traffic light control via multi-agent hierarchical reinforcement learning. In 19th international conference on automation science and engineering (CASE) 2023 (pp. 1-6). IEEE.
[21] Lim J, Dalmeijer K, Guhathakurta S, Van HP. The bicycle network improvement problem. Journal of Transportation Engineering, Part A: Systems. 2022; 148(11):04022095.
[22] Liu T, Meidani H. End-to-end heterogeneous graph neural networks for traffic assignment. Transportation Research Part C: Emerging Technologies. 2024; 165:1-14.
[23] Jia W, Ji M. Multi-agent deep reinforcement learning for large-scale traffic signal control with spatio-temporal attention mechanism. Applied Sciences. 2025; 15(15):1-16.
[24] Devailly FX, Larocque D, Charlin L. IG-RL: inductive graph reinforcement learning for massive-scale traffic signal control. IEEE Transactions on Intelligent Transportation Systems. 2021; 23(7):7496-507.
[25] Zhang G, Chang F, Jin J, Yang F, Huang H. Multi-objective deep reinforcement learning approach for adaptive traffic signal control system with concurrent optimization of safety, efficiency, and decarbonization at intersections. Accident Analysis & Prevention. 2024; 199:107451.
[26] Peng Z, Cheng H, Shi K, Zou C, Huang R, Li X, et al. Optimal tracking control for motion constrained robot systems via event-sampled critic learning. Expert Systems with Applications. 2023; 234:121085.
[27] Wang Y, Wu Y, Tang Y, Li Q, He H. Cooperative energy management and eco-driving of plug-in hybrid electric vehicle via multi-agent reinforcement learning. Applied Energy. 2023; 332:120563.
[28] Bideris-davos AA, Vovos PN. Co-optimization of power and water distribution systems for simultaneous hydropower generation and water pressure regulation. Energy Reports. 2024; 11:3135-48.
[29] Sun X, Lu Y, Wu Q, Li H, Xu Z, Zheng L, et al. An eco-driving strategy of vehicle comfort-safety and energy efficiency based on deep reinforcement learning. Alexandria Engineering Journal. 2025; 132:252-63.
[30] Oroojlooy A, Nazari M, Hajinezhad D, Silva J. Attendlight: universal attention-based reinforcement learning model for traffic signal control. Advances in Neural Information Processing Systems. 2020; 33:4079-90.