A reliable driving trajectory prediction of surrounding vehicles is an essential reference for decision-making and safe driving of an autonomous vehicle. Although predicting short-term trajectories can be well achieved, it is still very challenging for long-term prediction of trajectories since the prediction space grows exponentially. In this paper, we propose a novel architecture for trajectory prediction from factored intention estimation (I2T), which decouples the trajectory prediction space into a high-level space for intention estimation and a low-level space for motion prediction. The long-term dependencies between intention cues and future motions during driving are naturally extended to the internal sharing mechanism of I2T, leading to improved performance. Furthermore, we design a Prioriformer model to serve as the backbone network for I2T so that it can accurately capture the long-term dependency couplings related to the task of intention estimation or motion prediction. Prioriformer model adopts a personalized normalization method, which facilitates learning latent representations of long-term features and avoids getting stuck on local optimum. A designed multi-scale fusion encoder extracts features from various receptive fields and then learns richer information from the representation subspaces. An efficient non-autoregressive decoder reduces the pressure in long-term prediction of trajectories while avoiding cumulative errors. Experiments on three real-world motion datasets show that I2T can significantly outperform the state-of-the-art.

Various external factors that interfere with traffic flow, such as weather conditions, traffic accidents, incidents, and Points of Interest (POIs), need to be considered in performing traffic forecasting tasks. However, the current research methods encounter difficulties in effectively incorporating these factors with traffic characteristics and efficiently updating them, which leads to a lack of dynamics and interpretability. Moreover, capturing temporal dependence and spatial dependence separately and sequentially can result in issues, such as information loss and model errors. To address these challenges, we present a Knowledge Representation learning-actuated spatial–temporal graph neural network (KR-STGNN) for traffic flow prediction. We combine the knowledge embedding with the traffic features via Gated Feature Fusion Module (GFFM), and dynamically update the traffic features adaptively according to the importance of external factors. To conduct the co-capture of spatial–temporal dependencies, we subsequently propose a spatial–temporal feature synchronous capture module (ST-FSCM) combining dilation causal convolution with GRU. Experimental results on a real-world traffic data set demonstrate that KR-STGNN has superior forecasting performances over diverse prediction horizons, especially for short-term prediction. The ablation and perturbation analysis experiments further validate the effectiveness and robustness of the designed method.

Consensus is a common opinion evolution phenomenon in social networks. Existing studies generally believe that opinion consensus is easy to reach when the network is connected and only trust relationships exist among agents. Distinguishing from existing studies, this paper aims to explore how to reach opinion consensus on leader-follower signed social networks with both trust and distrust relationships among agents. With the help of theoretical tools including binary relation composition and products of nonnegative matrices, the inequality constraints for trust and distrust degrees are established, which ensures that trust relationships can counteract the negative effects of distrust relationships on opinion consensus. In addition, further analysis is conducted on the factors that affect the convergence rate of opinion consensus, including the longest positive distance from leader to the followers as well as the upper bound of communication delays. At last, numerical examples are provided to validate the results.

The unmanned aerial vehicles (UAVs)-assisted intelligent traffic perception system can provide effective situation awareness. However, UAVs are required to be recharged before the energy is exhausted, which may cause task interruption. To address this concern, the charging UAV (CUAV) is employed to provide wireless charging for the mission UAVs (MUAVs). This paper studies the charging scheduling problem of the CUAV under the premise of optimizing the MUAVs deployment. We first model the MUAVs deployment problem considering the energy consumption and data transmission and establish the CUAV charging model. Then, the above problem is formulated as a multi-objective multi-agent stochastic game process to simplify the decisions-making of MUAVs and CUAV, based on which we propose the utility-based Pareto optimal deployment and charging algorithm, which reduces the computing complexity by equivalent utility of the MUAVs while using Kullback-Leibler divergence to constrain solutions. Next, to ensure the effectiveness of policy update, the multi-agent communication protocol is adopted to improve policy exploration efficiency. Simulation results show that the proposed algorithm outperforms existing works in terms of energy efficiency and charging by comparing with the Pareto front of different methods, endurance anxiety of the MUAVs, and charging utilization under different task modes.

Multiplex networks have been widely used in information diffusion, social networks, transport, and biology multiomics. They contain multiple types of relations between nodes, in which each type of the relation is intuitively modeled as one layer. In the real world, the formation of a type of relations may only depend on some attribute elements of nodes. Most existing multiplex network embedding methods only focus on intralayer and interlayer structural information while neglecting this dependence between node attributes and the topology of each layer. Attributes that are irrelevant to the network structure could affect the embedding quality of multiplex networks. To address this problem, we propose a novel multiplex network embedding model with high‐order node dependence, called HMNE. HMNE simultaneously considers three properties: (1) intralayer high‐order proximity of nodes, (2) interlayer dependence in respect of nodes, and (3) the dependence between node attributes and the topology of each layer. In the intralayer embedding phase, we present a symmetric graph convolution‐deconvolution model to embed high‐order proximity information as the intralayer embedding of nodes in an unsupervised manner. In the interlayer embedding phase, we estimate the local structural complementarity of nodes as an embedding constraint of interlayer dependence. Through these two phases, we can achieve the disentangled representation of node attributes, which can be treated as fined‐grained semantic dependence on the topology of each layer. In the restructure phase of node attributes, we perform a linear fusion of attribute disentangled representations for each node as a reconstruction of original attributes. Extensive experiments have been conducted on six real‐world networks. The experimental results demonstrate that the proposed model outperforms the state‐of‐the‐art methods in cross‐domain link prediction and shared community detection tasks.

Social networks which are dynamic contain rich network structure and content information. In dynamic networks, it is necessary to discover communities and their topical meanings. However, existing methods either only discover communities with ignoring their topical meaning in dynamic networks, or they discover communities and their topics in static networks. In this paper, we identify the problem of dynamic topical community detection and propose a dynamic topical community detection(DTCD) method to detect communities and their topical meanings in dynamic networks. DTCD is a generative model integrating network structure, text and time. DTCD considers a community as a mixture of topics and generates the neighbors and documents of the node and their time stamps at the same time via the community. The latent variables are learned by collapsed Gibbs sampling. DTCD not only can find communities and their topics, but also capture the temporal variations of communities and topics. Experimental results on two real-world datasets demonstrate the effectiveness of DTCD.