Long-sequence traffic flow forecasting plays a crucial role in intelligent transportation systems. However, existing Transformer-based approaches face a quadratic complexity bottleneck in computation and are prone to over-smoothing in deep architectures. This results in overly averaged predictions that fail to capture the peaks and troughs of traffic flow. To address these issues, we propose a State-Space Generative Adversarial Network (SSGAN) with a state-space generator and a multi-scale convolutional discriminator. Specifically, a bidirectional Mamba-2 model was designed as the generator to leverage the linear complexity and efficient forecasting capability of state-space models for long-sequence modeling. Meanwhile, the discriminator incorporates a multi-scale convolutional structure to extract traffic features from the frequency domain, thereby capturing flow patterns across different scales, alleviating the over-smoothing issue and enhancing discriminative ability. Through adversarial training, the model is able to better approximate the true distribution of traffic flow. Experiments conducted on four real-world public traffic flow datasets demonstrate that the proposed method outperformed the baselines in both forecasting accuracy and computational efficiency.

Long-sequence traffic flow forecasting plays a crucial role in intelligent transportation systems. However, existing Transformer-based approaches face a quadratic complexity bottleneck in computation and are prone to over-smoothing in deep architectures. This results in overly averaged predictions that fail to capture the peaks and troughs of traffic flow. To address these issues, we propose a State-Space Generative Adversarial Network (SSGAN) with a state-space generator and a multi-scale convolutional discriminator. Specifically, a bidirectional Mamba-2 model is designed as the generator to leverage the linear complexity and efficient forecasting capability of state-space models for long-sequence modeling. Meanwhile, the discriminator incorporates a multi-scale convolutional structure to extract traffic features from the frequency domain, thereby capturing flow patterns across different scales, alleviating the over-smoothing issue, and enhancing discriminative ability. Through adversarial training, the model is able to better approximate the true distribution of traffic flow. Experiments conducted on four real-world public traffic flow datasets demonstrate that the proposed method outperforms baselines in both forecasting accuracy and computational efficiency.

Intercity mobility management plays a critical role in promoting urban sustainability and regional integration. Accurate prediction of mobility flows is essential for optimizing transport operations and enhancing resource allocation, especially during holidays. However, event-induced scarcity of reliable historical data makes it challenging for existing methods to capture the abrupt changes in holiday mobility patterns. To address these challenges, this article proposes the dynamic time-space holiday feature captor (DHC), which integrates spatiotemporal correlations in time-series data with cross-holiday similarities to effectively model the dynamic patterns of holiday mobility, thereby significantly improving prediction accuracy. The DHC model comprises four key modules: 1) vision dynamic extension (VDE); 2) spatiotemporal feature acquisition (SFA); 3) vision feature extraction (VFE); and 4) holiday feature fusion (HFF). The main experimental results show that DHC outperforms existing methods on multiple metrics, achieving reductions of 33.98%, 26.14%, and 35.75% in MAE, RMSE, and MAPE, respectively. Moreover, the model’s effectiveness and generalizability are further validated through ablation studies and additional experiments conducted across various holidays. This research advances mobility-flow prediction and management during holidays and contributes to the development of sustainable transportation systems.

Road traffic significantly contributes to fuel consumption and emissions. Fortunately, the advent of cooperative adaptive cruise control (CACC), facilitated by vehicle-to-vehicle (V2V) communication, reduces energy consumption and improves efficiency in transportation systems. Nevertheless, V2V communication performance (V2VCP) is highly vulnerable to degradation due to various factors. Limited comprehension exists regarding the generalized modeling of subpar V2V communication performance (SV2VCP), coupled with limited exploration of its resulting impacts on environmental sustainability. To bridge these gaps, this study presents the first attempt to assess the impact of SV2VCP on fuel consumption and exhaust emissions within the CACC framework. More specifically, we adopt the multi-predecessor following (MPF) topology and model SV2VCP scenarios, along with proposing five vehicle state update methods (VSUMs). Subsequently, by simulating various SV2VCP and driving scenarios, we comprehensively understand the effects of different VSUMs, SV2VCP, and abnormal vehicle positions on the safety, emissions, and energy consumption of the platoon. The results reveal that SV2VCP substantially impacts the fuel efficiency and emission performance of the CACC platoon, with fuel consumption during deceleration exceeding that of acceleration by approximately 14% when all vehicles are subject to SV2VCP. Furthermore, our study provides critical recommendations for optimal strategy selection, aiming to foster energy conservation and emission reductions, thereby promoting sustainable transport systems.