Spatially partitioning heterogeneous traffic networks into multiple subnetworks is crucial for practical tasks, such as distributed signal control and model parallel processing. Existing partitioning methods that account for traffic characteristics over a certain period typically require calculating similarities between the time series of all sensors. Due to the quadratic increase in complexity with network size, these methods are inefficient for large-scale networks and extended time periods. Additionally, calculating similarities requires complete data, making such methods highly sensitive to missing data and lacking robustness. To address these issues, this article proposes a four-step fundamental-diagram-informed traffic network partitioning method. First, spatially adjacent sensors are clustered into subclusters. Next, the S3 speed-occupancy function is used to fit the aggregated data of each subcluster to extract fundamental diagram information. Then, this information is used to perform secondary clustering to form clusters. Finally, the cluster boundaries are fine-tuned to produce subnetworks with smooth boundaries. The proposed method calculates the parameter similarity between subclusters instead of the time series similarity between all sensors. This reduces computational costs and effectively handles data missing. A case study using real-world data verifies the effectiveness of the proposed method and its stability in the presence of missing data. Compared to spectral clustering, the total within-cluster variance and NcutSilhouette metric decrease by 5.7% and 17.8%, respectively. The proposed method enhances distributed or parallel tasks on traffic networks by providing stable and meaningful partitioning results. This method is beneficial for the analysis and effective management of complex heterogeneous traffic networks.

Accurate prediction of spatial–temporal traffic flow drives innovation across various pertinent application domains, including traffic management and route planning. Graph Convolutional Neural Network (GCN) consistently assume a central role within forecasting frameworks. The effectiveness of GCN models significantly hinges on a well-constructed graph structure, whether explicitly defined or acquired through the training process. This structure establishes the mechanism through which messages are exchanged among diverse spatial locations. In the context of traffic flow data, both prior knowledge and unknown factors contribute to the graph structure. Considering both the information derived through algorithms (self-learned) and existing knowledge (predefined), which collectively shape the spatial–temporal patterns of traffic flow, we introduce a novel model named parallel self-learned and predefined joint spatial–temporal GCN (PSPJSTGCN) for traffic flow forecasting. Our model employs a gated mechanism to amalgamate predefined and self-learned graphs in parallel, enabling efficient extraction of spatial–temporal dependency information from both sources. Additionally, we leverage multiscale gated convolution to capture dynamic temporal dependencies across a wide range of receptive fields. We meticulously evaluate our proposed approach using four real-world datasets and substantiate its substantial superiority over prevailing state-of-the-art methods.

Traffic state prediction, a classical task for traffic management, is a central component of intelligent transport systems to maintain safe and efficient operation. While extensive and intensive research has been conducted on traffic state prediction, most studies have concentrated on enhancing the accuracy of specific traffic state parameters. However, traffic state is a co-evolutionary multivariate time series with various parameters such as flow, velocity, occupancy, etc. At the same time, traffic state data will inevitably be lost during collection. So accurate traffic prediction still faces the following challenges: First, how to deal with the complex missing situations in observational data? Second, how to learn the co-evolutionary relationships between different traffic state parameters while mining the high-dimensional spatio-temporal traffic state patterns? In this paper, we propose a mechanism-data blending-driven co-evolving traffic state parameter prediction method: multi-parameter hybrid tensor deep learning networks (MHT-Net), which consists of a multi-parameter tensor graph convolutional network (MTGCN) and a tensor recurrent neural network (T-RNN). MTGCN implements knowledge embedding of synergistic mechanisms between traffic parameters, ensuring that the road network spatial dependency and the synergistic influence relationship of the parameters can be obtained simultaneously; T-RNN is used to learn high-dimensional temporal features of traffic states. Experiment results on a real-world dataset from Jiangsu province outperform the state-of-the-art baselines, demonstrating the efficacy of the proposed method and providing an effective tool for traffic state prediction with missing values. A mechanism-data blending driven co-evolving traffic state parameter prediction method, multi-parameters hybrid tensor deep learning networks (MHT-Net) is proposed, which implements knowledge embedding of synergistic mechanisms between traffic parameters and learn the road network spatial dependency and the synergistic influence relationship of the parameters simultaneously. Experiment results demonstrate the efficacy of the proposed method and provide an effective tool for traffic state prediction with missing values.

Trajectory prediction is a crucial task for autonomous driving, but current models’ reliance on high-definition (HD) maps limits their broader applicability. To cope with this challenge, we propose a novel map-free trajectory prediction method that leverages spatiotemporal attention mechanisms. The method consists of three key stages: 1) we first encode spatial and temporal features separately using spatial and temporal attention mechanisms, 2) we then model spatial and temporal interactions through Crystal Graph Convolutional Networks (CGCN) and Multi-Head Attention (MHA), 3) finally, an adaptive anchor generation technique is introduced to tackle the multimodal trajectory prediction challenge. This self-adaptive technique generates context-specific anchors, enabling accurate prediction of multiple possible future vehicle trajectories. Extensive experiments on the Argoverse1 and V2X-Seq datasets validate the effectiveness of our approach. On the Argoverse1 dataset, our method outperforms CRAT-Pred by 5.8% in minADE and 6.25% in minFDE. On the V2X-Seq dataset, it achieves improvements of 82.6%, 85.1%, and 44.0% in minADE, minFDE, and MR, respectively, compared to the baseline model.

Deep Reinforcement Learning (DRL) has achieved significant advancements in the transportation domain, effectively enhancing traffic network efficiency, reducing pollutant emissions, and improving driving safety. A prominent approach within DRL, Hierarchical Reinforcement Learning (HRL), simplifies complex tasks by grouping states and decomposing the Markov Decision Process (MDP), facilitating the exploration in multi-dimensional state spaces. These concepts of state abstraction and temporal abstraction prove to be particularly beneficial in complex, high-risk transportation scenarios, such as on-ramp merging. In this context, this paper introduces a novel Multi-Option HRL (MO-HRL) framework with state segmentation. Unlike traditional option-based HRL, the proposed framework enables the simultaneous activation of multiple options, with each option observing diverse states. After carefully defining and justifying the framework, we apply MO-HRL to a simplified on-ramp merging scenario. To enhance training, curriculum learning is incorporated into the MO-HRL framework. Extensive experiments involve discussions of different training modes, the “shared critic” problem, and comparisons with state-of-the-art baselines. Additionally, a six-lane mainline on-ramp merging scenario, based on the NGSIM I-80 dataset, is constructed. Simulation results from both scenarios show that the proposed approach outperforms existing methods and maintains a balance between the mainline and on-ramp traffic.

Trajectory prediction plays a pivotal role in the autonomous driving systems. Existing methods generally employ agent-centric or scene-centric approaches to represent driving scenarios. However, these methods introduce significant redundant computations or pose losses, resulting in suboptimal prediction efficiency and accuracy. To tackle these problems, a novel multi-target trajectory prediction model, named Hierarchical Scene TRansformer (HSTR), is introduced. The driving scene is decomposed into two independent components by HSTR: global and local. In the global part, global interaction information is established and shared among all predicted agents, thereby reducing redundant computations. In the local part, an individual reference frame is established for each vehicle to eliminate the impact of pose variations and extract temporal features. Moreover, an adaptive anchor point generation method is proposed to address the challenge of capturing future modalities for vehicles. This method dynamically generates corresponding anchor points based on different driving scenarios to guide the prediction of trajectories across various modalities. The model performance is verified on the argoverse1 and argoverse2 datasets, and the experimental results demonstrate that competitive performance is achieved by HSTR in terms of efficiency and precision compared to the state-of-the-art methods.

On-ramp merging problem focuses on vehicle safety and traffic efficiency. It can be considered as a hierarchical planning scenario with free flow zone, preparation zone and merging zone. Previous researches consider Reinforcement Learning (RL) as a potential solution due to its general learning ability. However, flat-RL tends to consider the on-ramp merging problem as a whole, neglecting its hierarchical property and causing limited improvement. Instead, with temporal abstraction, Option-based Hierarchical Reinforcement Learning (HRL) is capable to solve complicated problem by using task decomposition, giving a hint to adapt various zones in on-ramp merging problem. We hence propose an HRL-based Macro-Micro Adaptive framework (HRL-MMA). In this end-to-end framework, a Macro-Micro Adaptive Layer (MMAL) provides both macroscopic traffic information and microscopic vehicle information to the framework. The macroscopic information aims to help the master of the framework to choose options of different capacities, while the latter guarantees the safety of merging. Extensive experiments involve both the state-of-the-art baselines and several variants of the proposed framework. Compared with the IDM model, the proposed HRL-MMA framework has a 46.98% increase on the network average velocity and a 59.16% improvement on the emergency braking rate, largely ameliorating the safety of the merging problem.

Car-following models have been widely applied and made remarkable achievements in traffic engineering. However, the traffic micro-simulation accuracy of car-following models in a platoon level, especially during traffic oscillations, still needs to be enhanced. Rather than using traditional individual car-following models, we proposed a new trajectory generation approach to generate platoon level trajectories given the first leading vehicle’s trajectory. In this article, we discussed the temporal and spatial error propagation issue for the traditional approach by a car following block diagram representation. Based on the analysis, we pointed out that error comes from the training method and the model structure. In order to fix that, we adopt two improvements on the basis of the traditional LSTM-based car-following model. We utilized a scheduled sampling technique during the training process to solve the error propagation in the temporal dimension. Furthermore, we developed a unidirectional interconnected LSTM model structure to extract trajectories features from the perspective of the platoon. As indicated by the systematic empirical experiments, the proposed novel structure could efficiently reduce the temporal-spatial error propagation. Compared with the traditional LSTM-based car-following model, the proposed model has almost 40% less error. The findings will benefit the design and analysis of micro-simulation for platoon-level car-following models.

Understanding the underlying patterns of the urban mobility dynamics is essential for both the traffic state estimation and management of urban facilities and services. Due to the coupling relationship of generative factors in spatial-temporal domain, it is challenging to model the citywide traffic dynamics under a structural pattern of critical features such as hours of days, days of weeks and weather conditions. To address this challenge, this article develops a disentangled representation learning framework to learn an interpretable factorized representation of the independent data generative factors. In order to make full use of the knowledge on generative factors, this article proposes spatial-temporal generative adversarial network (ST-GAN) to assign the generative factors of traffic flow to the feature vector in latent space and reconstructs the high-dimensional citywide traffic flow from the given factors. With the help of the disentangled representations, the decomposed feature vector in latent space discloses the relationship between underlying patterns and citywide traffic dynamics. Several comprehensively experiments show that ST-GAN not only effectively improves the prediction accuracy but also promisingly characterize structural properties of the traffic evolution process.

Modern automotive systems have been equipped with a highly increasing number of onboard computer vision hardware and software, which are considered to be beneficial for achieving eco-driving. This article combines computer vision and deep reinforcement learning (DRL) to improve the fuel economy of hybrid electric vehicles. The proposed method is capable of autonomously learning the optimal control policy from visual inputs. The state-of-the-art convolutional neural networks-based object detection method is utilized to extract available visual information from onboard cameras. The detected visual information is used as a state input for a continuous DRL model to output energy management strategies. To evaluate the proposed method, we construct 100 km real city and highway driving cycles, in which visual information is incorporated. The results show that the DRL-based system with visual information consumes 4.3-8.8% less fuel compared with the one without visual information, and the proposed method achieves 96.5% fuel economy of the global optimum-dynamic programming.

Accurate and robust feature matching is a critical issue in the preprocessing of multispectral image data sets. Higher order features, such as lines can provide useful matching information but are heavily affected by the unreliable detection of lines. Existing methods typically make the unrealistic assumption that end points of lines can be accurately detected across the reference and test images. To address the unreliable detection of line end points, this paper proposes mapping line intersections and then employing tentatively mapped intersections as “anchor” points to compute line descriptors. The computed line descriptors are utilized to determine whether the two lines forming an intersection are matched with the two lines forming its mapped intersection. This eliminates the reliance on the accurate detection of line end points and results in improved matching accuracy. The proposed method is tested on a large number of multispectral images containing various scenes. Experimental results show that it can effectively deal with the detection inaccuracy of end points for line matching.

Low-rank tensor completion methods have been advanced recently for modeling sparsely observed data with a multimode structure. However, low-rank priors may fail to interpret the model factors of general tensor objects. The most common method to address this drawback is to use regularizations together with the low-rank priors. However, due to the complex nature and diverse characteristics of real-world multiway data, the use of a single or a few regularizations remains far from efficient, and there are limited systematic experimental reports on the advantages of these regularizations for tensor completion. To fill these gaps, we propose a modified CP tensor factorization framework that fuses the l2norm constraint, sparseness (l1norm), manifold, and smooth information simultaneously. The factorization problem is addressed through a combination of Nesterov's optimal gradient descent method and block coordinate descent. Here, we construct a smooth approximation to the l1norm and TV norm regularizations, and then, the tensor factor is updated using the projected gradient method, where the step size is determined by the Lipschitz constant. Extensive experiments on simulation data, visual data completion, intelligent transportation systems, and GPS data of user involvement are conducted, and the efficiency of our method is confirmed by the results. Moreover, the obtained results reveal the characteristics of these commonly used regularizations for tensor completion in a certain sense and give experimental guidance concerning how to use them.

Clustering is among the most fundamental tasks in machine learning and artificial intelligence. In this paper, we propose Variational Deep Embedding (VaDE), a novel unsupervised generative clustering approach within the framework of Variational Auto-Encoder (VAE). Specifically, VaDE models the data generative procedure with a Gaussian Mixture Model (GMM) and a deep neural network (DNN): 1) the GMM picks a cluster; 2) from which a latent embedding is generated; 3) then the DNN decodes the latent embedding into an observable. Inference in VaDE is done in a variational way: a different DNN is used to encode observables to latent embeddings, so that the evidence lower bound (ELBO) can be optimized using the Stochastic Gradient Variational Bayes (SGVB) estimator and the reparameterization trick. Quantitative comparisons with strong baselines are included in this paper, and experimental results show that VaDE significantly outperforms the state-of-the-art clustering methods on 5 benchmarks from various modalities. Moreover, by VaDE's generative nature, we show its capability of generating highly realistic samples for any specified cluster, without using supervised information during training.

Short-term traffic prediction plays a critical role in many important applications of intelligent transportation systems such as traffic congestion control and smart routing, and numerous methods have been proposed to address this issue in the literature. However, most, if not all, of them suffer from the inability to fully use the rich information in traffic data. In this paper, we present a novel short-term traffic flow prediction approach based on dynamic tensor completion (DTC), in which the traffic data are represented as a dynamic tensor pattern, which is able capture more information of traffic flow than traditional methods, namely, temporal variabilities, spatial characteristics, and multimode periodicity. A DTC algorithm is designed to use the multimode information to forecast traffic flow with a low-rank constraint. The proposed method is evaluated on real-world data sets and compared with other state-of-the-art methods, and the efficacy of the proposed approach is validated on the experiments of traffic flow prediction, particularly when dealing with incomplete traffic data.

The tensor completion problem is to recover a low-n-rank tensor from a subset of its entries. The main solution strategy has been based on the extensions of trace norm for the minimization of tensor rank via convex optimization. This strategy bears the computational cost required by the singular value decomposition (SVD) which becomes increasingly expensive as the size of the underlying tensor increase. In order to reduce the computational cost, we propose a multi-linear low-n-rank factorization model and apply the nonlinear Gauss–Seidal method that only requires solving a linear least squares problem per iteration to solve this model. Numerical results show that the proposed algorithm can reliably solve a wide range of problems at least several times faster than the trace norm minimization algorithm.

The problem of recovering data in multi-way arrays (i.e., tensors) arises in many fields such as image processing and computer vision, etc. In this paper, we present a novel method based on multi-linear n-rank and ℓ0norm optimization for recovering a low-n-rank tensor with an unknown fraction of its elements being arbitrarily corrupted. In the new method, the n-rank and ℓ0norm of the each mode of the given tensor are combined by weighted parameters as the objective function. In order to avoid relaxing the observed tensor into penalty terms, which may cause less accuracy problem, the minimization problem along each mode is accomplished by applying the augmented Lagrange multiplier method. The proposed approach is evaluated both on simulated data and real world data. Experimental results show that our proposed method tends to deliver higher-quality solutions with faster convergence rate compared with previous methods.

How to teach computer to classify pictures into different emotional categories automatically as humans? It is a quite interesting and challenging research direction. This paper defines the latent emotional semantic factors and proposes a novel approach for emotional picture categorization. Unlike traditional methods, the latent emotional semantic factor is defined for representing the middle level semantic concept, and it can effectively bridge the “semantic gap”. In the experiments made on the International Affective Picture System (IAPS), the proposed approach for emotional picture categorization distinctly outperforms the latest method.

The first Chinese unmanned ground vehicles competition - The 2009 Future Challenge: Intelligent Vehicles and Beyond (FC'09) pushed China's unmanned vehicles out of laboratories and into application environments. In order to further promote the development of unmanned vehicle technologies, the test and evaluation system for unmanned vehicles needs to be studied. The design method of test environment is proposed in accordance with the definition and classification of test environment elements. Based on the multi-platform and multi-sensor, an omnidirectional video monitoring test system of unmanned vehicles is built. The fuzzy comprehensive evaluation method combined with AHP (analytic hierarchy process) is applied to the comprehensive evaluation of unmanned vehicles. The evaluation examples of unmanned vehicles show that the proposed evaluation system can quantitatively evaluate the overall technical performance and individual technical performance of unmanned vehicles.