Next parking recommendation improves the utilization of existing parking resources, reduces driver uncertainty at trip endpoints and enhances travel experience. However, existing approaches to parking often overlook the joint modeling of spatial dependencies between facilities and temporal regularities in user mobility. To address this problem, we propose a parking recommendation model named ParkPredictNet that combines a graph embedding module to capture spatial relations among parking facilities, a Transformer encoder to learn sequential dependencies in parking trajectories, and multiple temporal encoding to represent periodicity and parking event intervals. This design enables the joint modeling of spatial and fine-grained temporal patterns to power personalized top-k recommendations. On a real-world dataset, experimental results show ParkPredictNet outperforms baseline methods across both accuracy and ranking metrics. Comprehensive ablations valuate the contribution of each component. The system targets real-world deployment, delivering low-latency, personalized recommendations to many drivers in parallel.

Along with the proliferation of electric vehicles (EVs), optimizing the use of EV charging space can significantly alleviate the growing load on intelligent transportation systems. As the foundation to achieve such an optimization, a spatiotemporal method for EV charging demand prediction in urban areas is required. Although several solutions have been proposed by using data-driven deep learning methods, it can be that these performance-oriented approaches may struggle to correctly understand the underlying factors influencing charging demand, particularly charging prices. A representative case that highlights the challenge faced by existing methods is their potential misinterpretation of high prices during peak times, leading to an incorrect assumption that higher prices correspond to increased demand. To address the challenges associated with training an accurate and reliable prediction model for EV charging demand, this paper proposes a novel approach called PAG, which leverages the integration of graph and temporal attention mechanisms for effective feature extraction and introduces physics-informed meta-learning in the pre-training step to facilitate prior knowledge learning. Evaluation results on a dataset of 18,061 EV charging piles in Shenzhen, China, show that the proposed approach can achieve state-of-the-art forecasting performance and the ability to understand the adaptive changes in charging demands caused by price fluctuations.

Parking availability prediction is an important part of urban intelligent transportation systems (ITS), which allows drivers to be informed of the parking availability in advance, thereby reducing cruising for parking time and reducing pollution emissions. However, existing studies of parking availability prediction rarely incorporate the effect of price factors and cannot be trained to obtain the price elasticity of parking demand. In this paper, we propose a neural network-based approach called physics-informed parking (PI-parking). The model employs physics-informed neural networks (PINNs) to add price information into model training, which not only achieves the parking availability prediction, but also calculates the price elasticity of demand of different parking lots during the training process. On the SF Park project, experiment results show that our model outperforms 17.711% on average than compared models. For price elasticity of parking demand, we find that the values are distributed between 0.02 and 0.31, and most of the values are between 0.25 and 0.27. In addition, there is a similarity in the elasticity values for adjacent neighborhoods.

Traffic spatial-temporal prediction is essential for intelligent transportation systems. However, the current approach relies heavily on expert knowledge and time-consuming manual modeling. Neural architecture search can build models adaptively, but it is rarely used for traffic spatial-temporal prediction, nor is it designed specifically for traffic spatial-temporal feature. In response to the above problems, we propose neural architecture search spatial-temporal prediction (NASST), which is a method to automatically generate a traffic spatial-temporal prediction network by performing a differentiable neural network architecture search in an optimized search space. First, we adopt a differentiable neural architecture search method to continuously relax the discrete traffic spatial-temporal prediction model architecture search, and adopt a fusion strategy of comprehensive concatenate and addition (CA) to achieve efficient neural architecture search. Second, we optimize the search space and introduce a series of classic traffic spatial-temporal feature extraction modules, which are more in line with the architectural requirements of traffic spatial-temporal prediction network. Finally, our model is validated on two public traffic datasets and achieves the best predictions. Compared with traditional manual modeling methods, our method can realize the automatic search of high-precision predictive model architectures, which improves the modeling efficiency.

In a city, the usage optimization of parking spaces with a near real-time response to car drivers can significantly reduce the unnecessary cruising for parking and the additional congestion of regional traffic. As the foundation to achieve such an optimization, a parking occupancy prediction method is required to address the emerging challenges of training a simple but effective model. To fill the gap, this paper proposes a novel approach that enables the integration of Time Series Decomposition (TSD), Gated Recurrent Unit (GRU), and First-order Model-agnostic Meta-learning (FOMAML) for feature engineering, model building, and model pre-training, respectively. Moreover, as shown by a detailed evaluation, such an integration strengthens the proposed approach, named Meta TSD-GRU, which outperforms other state-of-the-art methods with 1) prediction errors reduced by about 45% on average, 2) the speed of model adaptation and convergence improved about 2 and 102 times against the methods with and without pre-training, respectively, and 3) the generalizability of the model enhanced to handle various time intervals of forecasting and types of parking lots under a consistent and stable performance.