@Article{cmc.2026.084267,
AUTHOR = {Yunrui Bi, Qiliang Yang, Qinglin Ding, Bin Ran, Kun Liu, Mingjie Zhang},
TITLE = {A Hybrid Bio-inspired Type-2 Fuzzy Reinforcement Learning Framework for Regional Traffic Signal Coordination Control},
JOURNAL = {Computers, Materials \& Continua},
VOLUME = {},
YEAR = {},
NUMBER = {},
PAGES = {{pages}},
URL = {http://www.techscience.com/cmc/online/detail/27709},
ISSN = {1546-2226},
ABSTRACT = {To improve regional traffic signal coordination under uncertain and dynamic traffic conditions, this paper proposes a hybrid Type-2 fuzzy reinforcement learning framework integrated with Beetle Antennae Search (BAS) and Deep Q-Network (DQN), named Type-2 fuzzy Beetle Antennae Search and Deep Q-Network (T2-BAS-DQN). In this framework, DQN remains active during online signal control, while the Type-2 fuzzy module provides uncertainty-aware correction for phase selection and green-time adjustment. BAS is used only in the offline training stage to optimize a low-dimensional parameter vector related to fuzzy correction, reward adjustment, and coordination pressure. A 3 <mml:math id="mml-ieqn-1"><mml:mo>×</mml:mo></mml:math> 3 Simulation of Urban MObility (SUMO) road network is constructed to compare the proposed method with Fixed-Time, Max-Pressure, CoLight, DQN, and related baselines. The results show that T2-BAS-DQN achieves lower queue length, shorter waiting time, and higher average speed under different traffic demand levels. In addition, a real-geometry SUMO simulation calibrated with field traffic counts from Jiangning District, Nanjing, is conducted to further examine its applicability. Compared with DQN in this calibrated scenario, T2-BAS-DQN reduces average queue length by 47.4%, decreases average waiting time by 24.4%, and increases average travel speed by 11.0%. These results indicate that the proposed framework provides competitive simulation performance and improves several traffic operation indicators in the tested scenarios.},
DOI = {10.32604/cmc.2026.084267}
}