# CROSS: A Mixture-of-Experts Reinforcement Learning Framework for Generalizable Large-Scale Traffic Signal Control Source: [CROSS: A Mixture-of-Experts Reinforcement Learning Framework for Generalizable Large-Scale Traffic Signal Control]() Collection index: [llms.txt]() Data updated: 2026-03-28T05:00:04.914Z Authors: Xibei Chen, Yifeng Zhang, Yuxiang Xiao, Mingfeng Fan, Maonan Wang, Guillaume Sartoretti Published: 2026-03-26T01:48:15.000Z ## Abstract Recent advances in robotics, automation, and artificial intelligence have enabled urban traffic systems to operate with increasing autonomy towards future smart cities, powered in part by the development of adaptive traffic signal control (ATSC), which dynamically optimizes signal phases to mitigate congestion and optimize traffic. However, achieving effective and generalizable large-scale ATSC remains a significant challenge due to the diverse intersection topologies and highly dynamic, complex traffic demand patterns across the network. Existing RL-based methods typically use a single shared policy for all scenarios, whose limited representational capacity makes it difficult to capture diverse traffic dynamics and generalize to unseen environments. To address these challenges, we propose CROSS, a novel Mixture-of-Experts (MoE)-based decentralized RL framework for generalizable ATSC. We first introduce a Predictive Contrastive Clustering (PCC) module that forecasts short-term state transitions to identify latent traffic patterns, followed by clustering and contrastive learning to enhance pattern-level representation. We further design a Scenario-Adaptive MoE module that augments a shared policy with multiple experts, thus enabling adaptive specialization and more flexible scenario-specific strategies. We conduct extensive experiments in the SUMO simulator on both synthetic and real-world traffic datasets. Compared with state-of-the-art baselines, CROSS achieves superior performance and generalization through improved representation of diverse traffic scenarios. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Traffic Signal Control on Synthetic Arterial 4x4 Medium](): Traffic Signal Control; Synthetic Arterial 4x4 Medium; Queue Length: 0.95. - [Traffic Signal Control on Synthetic Grid 5x5 Hard](): Traffic Signal Control; Synthetic Grid 5x5 Hard; Queue Length: 1.05. - [Traffic Signal Control on Jinan Hard Real-World](): Traffic Signal Control; Jinan Hard Real-World; Average Queue Length: 0.65. - [Traffic Signal Control on Jinan Medium Real-World 2](): Traffic Signal Control; Jinan Medium Real-World 2; Queue Length: 0.38. - [Traffic Signal Control on Jinan Hard, Real-World 3](): Traffic Signal Control; Jinan Hard, Real-World 3; Queue Length: 0.51. - [Traffic Signal Control on Hangzhou Medium, Real-World 1 (test)](): Traffic Signal Control; Hangzhou Medium, Real-World 1 (test); Average Queue Length: 0.17. - [Traffic Signal Control on Hangzhou Hard, Real-World 1 (test)](): Traffic Signal Control; Hangzhou Hard, Real-World 1 (test); Average Queue Length: 0.42. - [Traffic Signal Control on Synthetic Grid 4x4 Easy](): Traffic Signal Control; Synthetic Grid 4x4 Easy; Queue Length: 0.05. ## Original sources and implementation [Original paper]() [PDF]()