Sensing-assisted beamforming techniques, with the aid of multimodal fusion perception, ensure highly reliable beam selection for V2I communication. However, due to the frequent communication path updates in high-mobility scenarios and the limited computing resources of base stations, the high-burden multimodal fusion computation make communication delays unavoidable. In this paper, we propose BeMamba, a novel multimodal fusion framework based on state space model for beamforming to balance the reliability and low latency of communication. Benefiting from the hidden state’s efficient sequence modeling ability with linear computational complexity, we designTime Sequence MambaandModal Sequence Mambato achieve intra-modal temporal fusion and cross-modal feature fusion. In addition, we develop dedicated data pre-processing methods as well as modality-specific feature extractors for the accessible modalities: image, LiDAR, radar, and GPS. On the DeepSense6G benchmark, our method achieves a 5.16% improvement in beam prediction accuracy, a 77.88% reduction in computational load, and a 4.56 times increase in inference speed.

Deep learning methods have achieved state-of-the-art performance in industrial fault diagnosis within the supervised learning paradigm. However, annotated data are scarce in industry, which can lead to overfitting and hinder their application. To address this issue, this article proposes a semisupervised learning framework that leverages self-supervised learning on abundant unlabeled data and supervised learning on limited labeled data simultaneously. Self-supervised learning captures inherent evolutionary dynamics, while supervised learning focuses on discriminative features. Specifically, a cross-prediction task on two augmented views of unlabeled data is devised using contextual representation. These contextual representations are used to construct a relational graph of unlabeled samples, which is then aligned with the corresponding logits graph. By facilitating interactions between the two tasks, the proposed framework achieves efficient fault diagnosis. Experiments on the Tennessee Eastman process and three-phase flow Facility datasets demonstrate the superiority of the proposed framework over other label-efficient methods.

Time-series anomaly detection plays a crucial role in industrial fault detection. Most existing studies follow either an unsupervised setting, which is prone to false alarms, or a supervised setting, which is time-consuming and labor-intensive. To address these limitations, we adopt an innovative weakly supervised paradigm for industrial fault detection, where segment-level labels are provided during training, while point-level predictions are made during inference. Within this paradigm, we propose an innovative$C$-ary tree-based multi-instance learning (MIL) framework. First, the entire time series is represented as a$C$-ary tree, where nodes representing subsequences of different lengths are treated as instances in the MIL framework. This design allows for the detection of both point and collective anomalies. Second, to detect out-of-distribution (OOD) anomalies that are not visible during training, we develop a vector quantization module to memorize regular historical patterns. OOD anomalies are then detected when they show significant discrepancies from all memorized patterns. Finally, we enhance the MIL framework with an attention-based pooling mechanism that allocates greater focus on anomalous instances, further improving detection performance. To validate the effectiveness of our method, we conduct experiments on four real-world industrial time-series datasets. The results show that our method outperforms existing approaches by at least 6.01% in AUROC under weak supervision.

Trajectory prediction is an essential component in autonomous driving, particularly for collision avoidance systems. Considering the inherent uncertainty of the task, numerous studies have utilized generative models to produce multiple plausible future trajectories for each agent. However, most of them suffer from limited representation ability or unstable training issues. To overcome these limitations, we propose utilizing the diffusion model to generate the distribution of future trajectories. Two cruxes are to be settled to realize such an idea. First, the diversity of intention is intertwined with the uncertain surroundings, making the true distribution hard to parameterize. Second, the diffusion process is time-consuming during the inference phase, rendering it unrealistic to implement in a real-time driving system. We propose an Intention-aware denoising Diffusion Model (IDM), which addresses the above two problems. We decouple the original uncertainty into intention uncertainty and action uncertainty and model them with two dependent diffusion processes. To decrease the inference time, we reduce the variable dimensions in the intention-aware diffusion process and restrict the initial distribution of the action-aware diffusion process, which leads to fewer diffusion steps. To validate our approach, we conduct experiments on the Stanford Drone Dataset (SDD) and the ETH/UCY dataset. Our methods achieve state-of-the-art results, with a minFDE of 13.83 pixels on the SDD dataset and 0.36 meters on ETH/UCY datasets. Compared with the original diffusion model, IDM reduces inference time by two-thirds. Interestingly, our experiments further reveal that introducing intention information is beneficial in modeling the diffusion process of fewer steps.

Semantic communication (SemCom) has emerged as a key technology for the forthcoming sixth-generation (6G) network, attributed to its enhanced communication efficiency and robustness against channel noise. However, the open nature of wireless channels makes them vulnerable to eavesdropping, which poses a serious threat to privacy. To address this issue, we propose a novel secure semantic communication (SemCom) approach for image transmission, which integrates steganography technology to conceal private information within non-private images (host images). Specifically, we propose an invertible neural network (INN)-based signal steganography approach that embeds channel input signals of a private image into those of a host image before transmission. This ensures that the original private image can be reconstructed from the received signals at the legitimate receiver, while the eavesdropper can only decode the information of the host image. Simulation results demonstrate that the proposed approach maintains comparable reconstruction quality of both host and private images at the legitimate receiver, compared to scenarios without any secure mechanisms. Moreover, the results indicate that the eavesdropper is only able to reconstruct host images, showcasing the enhanced security provided by our approach.

Time series anomaly detection (TSAD) plays a vital role in various domains such as healthcare, networks and industry. Considering labels are crucial for detection but difficult to obtain, we turn to TSAD with inexact supervision: only series-level labels are provided during the training phase, while point-level anomalies are predicted during the testing phase. Previous works follow a traditional multi-instance learning (MIL) approach, which focuses on encouraging high anomaly scores at individual time steps. However, time series anomalies are not only limited to individual point anomalies, they can also be collective anomalies, typically exhibiting abnormal patterns over subsequences. To address the challenge of collective anomalies, in this paper, we propose a tree-based MIL framework (TreeMIL). We first adopt an N-ary tree structure to divide the entire series into multiple nodes, where nodes at different levels represent subsequences with different lengths. Then, the subsequences’ features are extracted to determine the presence of collective anomalies. Finally, we calculate point-level anomaly scores by aggregating features from nodes at different levels. Experiments conducted on seven public datasets and eight baselines demonstrate that TreeMIL achieves an average 32.3% improvement in F1-score compared to previous state-of-the-art methods. The code is available at https://github.com/fly-orange/TreeMIL.

Artificial neural networks (ANNs) have been widely used for tasks like electricity theft detection (ETD) in smart meters. However, due to the subtle mechanisms and inherent opaque characteristics, ANNs are vulnerable to attacks. Although this attack surface poses significant risks, it has been largely overlooked in industrial scenarios. To alert the widespread adoption of industrial intelligence, this article studies the impact of backdoor attacks in ETD for the first time and proposes a feature attention distillation (FAD) defense. First, the attack surface in current model training pipeline is analyzed, and the adversaries can embed malicious backdoors for specific triggers to escape ETD. Then, six prevalent ANN-based models are tested and the adversaries can bypass the backdoored ETD models with success rates over 90.53%, which would inevitably bring huge losses to electricity companies. We further argue that the electricity companies can mitigate such attacks when noticing the abnormal nontechnical loss. A novel FAD defense that aligns the intermediate feature maps between fine-tuned and backdoored models is proposed, which can eliminate backdoors more efficiently with few resources compared with two classic defenses. The average attack success rate can drop by 90.71% with slight impacts on ETD performance. This work sheds light on a novel but perilous attack surface, and raises a warning for the wide adoption of artificial intelligence in smart measurement scenarios.

Traffic incident detection is a critical task within traffic monitoring systems, enabling on-the-fly alerts for emergency actions. Numerous efforts have been made to detect and localize traffic incidents using data recorded by inductive loop detectors. However, they only focus on the node-level incidents that happen within the surveillance areas and ignore the edge-level ones that take place outside of these areas. In this paper, we propose to detect both kinds of incidents simultaneously based on the sparsely distributed sensors. An important challenge is how to explicitly model the edge status and detect this kind of incidents. Additionally, capturing complex relationships among traffic dynamics, road locations, and temporal information is non-trivial. In this paper, we first describe the traffic dynamics by a fine-grained graph where the sensor range is designed as a hyper-parameter to control the coverage boundaries. Then, we propose an Edge-and Node-aware Dual AutoEncoder (ENDAE), where the correlations are decoupled into inter-nodes, inter-series and inter-attribute parts, which are further captured via node encoder, temporal encoder and attribute encoder, respectively. Furthermore, the reconstruction errors are calculated for node-level and edge-level event detection separately. The overall method is evaluated based on two real-world datasets from Bay Area and Los Angeles in California. ENDAE surpasses all the state-of-the-art method in both kinds of incidents, with at least 12.5% improvement in recall and 18.5% decrease in delay. Notably, for edge-level incidents, ENDAE achieves double the recall of the previous SOTA methods.

Wind speed forecasting provides the upcoming wind information and is important to the safe operation of High-Speed Railway (HSR). However, it remains a challenge due to the stochastic and highly varying characteristics of wind. In this paper, we propose a novel Transformer-based method for short-term wind speed forecasting, named WindTrans. Two major cruxes are addressed. First, the task is performed on fine-grained wind speed gathered from multiple sensors. These data present dynamic intra-series and inter-series correlations, which are hard for previous methods to recover. We advance a Transformer-based deep learning model, which has two distinctive characteristics: (1) a graph encoder, which captures the dynamic spatial correlation among wind speeds at different locations, and (2) a temporal decoder to model long sequence wind speed time series, which is resistant to noise in time series. Second, wind speed patterns gradually evolve in long-term periods, thus deactivating prediction models trained on historical data. To tackle this bottleneck, we put forward an experience replay-based scheme to renew the model regularly. To ensure that the renewed model still dominates historical wind patterns, we store and replay only a small portion of historical data named episodic memory. A simple but efficient strategy is designed to constitute episodic memory and thus relieve the computation burden. Experiments conducted on two real-world datasets demonstrate the superiority of our method over existing approaches. Particularly, WindTrans surpasses state-of-the-art methods by up to 36.7%, 29.3% and 13.3% improvement in MAPE measure for 1 hour ahead prediction on 10-minute, 5-minute, and 1-minute-based tasks, respectively. Furthermore, via our continual learning scheme, the model retains competitive performance with only 6.9% datum stored and retrained on.

Running at a fast speed, the high-speed train is prone to be interrupted by the surrounding strong wind. To ensure the safety of the trains, an effective approach is to deploy anemometers alongside the railway, such that the real-time and short-term predicted wind speed can be reported, and be further used by dispatchers to take protective actions in advance. However, in certain situations, the solely predicted wind speed is not informative enough to describe the wind status. It is difficult to tell if a strong wind incident could happen when the predicted wind speed is slightly lower than the strong wind threshold. We take the first attempt to predict the strong wind risk alongside the high-speed railway (HSR). A new model, called Multiple Attention Layer based Multi-Instance Learning (MAL-MIL), is proposed to address this problem. The key idea is to estimate the possibility that the actual wind speed exceeds the threshold conditionally on the predicted wind status. Based on attention mechanisms and long-short term memory network, the model can firstly generate deep representations of the future wind status. Then, though there is a lack of the risk ground truth, the multi-instance learning process facilitates the training procedure so that the relationships between these deep representations and the strong wind incidents could be quantified. Furthermore, considering the practicality of the model, we also design a result justification module to explain the reported risk. The superior performance is finally verified based on a real-world dataset.