Enhancing the accuracy of traffic prediction relies on building a graph that effectively captures the intricate spatiotemporal correlations in traffic data. It is a widely observed phenomenon that different urban traffic activities exhibit an asymmetric mutual influence. However, existing methods for graph construction largely overlook this characteristic. To bolster prediction performance, this article introduces an attention mechanism based on transfer entropy (TE) to quantify the complex and asymmetric spatiotemporal correlations among traffic data. Subsequently, a graph attention network based on simplified TE calculation is devised to construct directed spatiotemporal graphs from heterogeneous traffic data. Finally, a directed spatiotemporal graph neural network is employed for training and prediction purposes. Experimental results demonstrate that our approach surpasses existing mainstream methods for spatiotemporal graph construction, leading to significant improvements in predicting heterogeneous traffic data. Further analysis reveals that TE exhibits higher sensitivity to asymmetric spatiotemporal influences in heterogeneous traffic environments compared to commonly used data dependency inference algorithms. This finding further validates the feasibility and advancement of our method in predicting heterogeneous traffic data.

Spread of parking difficulty can be regarded as a special cascading failure process of urban parking systems. A comprehensive understanding of this process can be greatly helpful to build a more robust parking system. Parking network, a specified complex network, is proposed to model, simulate, and analyse the failure process of urban parking systems in this paper. This model is applied to the analysis of parking systems in an abstract city grid and the downtown area of Luohu, Shenzhen. The results demonstrate that the parking network can capture subtle variations among various parking cruising behaviours or strategies from a network perspective. To enhance the utility of the parking network, an auxiliary indicator named “Parking Difficulty Index” is introduced to help assess the failure degree of urban parking system, estimate the optimal timing for parking guidance intervention, and evaluate the effectiveness of various guidance strategies in mitigating parking difficulties.

With the sharp increase in the number of vehicles, the issue of parking difficulties has emerged as an urgent challenge that many cities need to address promptly. In the task of predicting large-scale urban parking data, existing research often lacks effective deep learning models and strategies. To tackle this challenge, this article proposes an innovative framework for predicting large-scale urban parking graphs leveraging real-time service capabilities, aimed at improving the accuracy and efficiency of parking predictions. Specifically, we introduce a graph attention mechanism that assesses the real-time service capabilities of parking lots to construct a dynamic parking graph that accurately reflects real preferences in parking behavior. To effectively handle large-scale parking data, this study combines graph coarsening techniques with temporal convolutional autoencoders to achieve unified dimension reduction of the complex urban parking graph structure and features. Subsequently, we use a spatiotemporal graph convolutional model to make predictions based on the coarsened graph, and a pretrained autoencoder-decoder module restores the predicted results to their original data dimensions, completing the task. Our methodology has been rigorously tested on a real dataset from parking lots in Shenzhen. The experimental results indicate that compared to traditional parking prediction models, our framework achieves improvements of 46.8% and 30.5% in accuracy and efficiency, respectively. Remarkably, with the expansion of the graph’s scale, our framework’s advantages become even more apparent, showcasing its substantial potential for solving complex urban parking dilemmas in practical scenarios.

Autonomous vehicles (AVs) rely on LiDAR and camera sensors to perceive their environment. However, adverse weather conditions, such as rain, snow, and fog, negatively affect these sensors, reducing their reliability by introducing unwanted noise. Effective denoising of multimodal sensor data is crucial for safe and reliable AV operation in such circumstances. Existing denoising methods primarily focus on unimodal approaches, addressing noise in individual modalities without fully leveraging the complementary nature of LiDAR and camera data. To enhance multimodal perception in adverse weather, we propose a novel Adaptive Cross-Modal Denoising (ACMD) framework, which leverages modality-specific self-denoising encoders, followed by an Adaptive Bridge Controller (ABC) to evaluate residual noise and guide the direction of cross-modal denoising. Following this, the Cross-Modal Denoising (CMD) module is introduced, which selectively refines the noisier modality using semantic guidance from the cleaner modality. Synthetic noise was added to both sensors’ data during training to simulate real-world noisy conditions. Experiments on the WeatherKITTI dataset show that ACMD surpasses traditional unimodal denoising methods (Restormer, PathNet, BM3D, PointCleanNet) by 28.2% in PSNR and 33.3% in CD, and outperforms state-of-the-art fusion models by 16.2% in JDE. The ACMD framework enhances AV reliability in adverse weather conditions, supporting safe autonomous driving.

Coverage model is often applied in spatial networks to evaluate site planning and robustness of networks. The visualized coverage model is helpful for people to understand and analyze these spatial data, in support for better decision. In this paper, we propose a novel dynamic coverage model for carparks around the city. Every carpark is quantified dynamically by its service capability, which is calculated in real time by the algorithm ParkingRank proposed in our previous work. Then an improved weighted Voronoi diagram featuring boundary constraint is introduced to map the quantified service capability of each carpark to the geospatial space, forming a seamless and non-overlapping coverage over the city. The best advantage of this model is the coverage scope of each carpark will dynamically adjust with its service capability varying, while the seamless and non-overlapping feature of coverage would be kept throughout from the perspective of the whole city. It is the first time that the network of carparks is modeled in a visual way, from which the city-wide parking management will get powerful decision supports in the future.