This work introduces BEV-LIO(LC), a novel LiDAR-Inertial Odometry (LIO) framework that combines Bird’s Eye View (BEV) image representations of LiDAR data with geometry-based point cloud registration and incorporates loop closure (LC) through BEV image features. By normalizing point density, we project LiDAR point clouds into BEV images, thereby enabling efficient feature extraction and matching. A lightweight convolutional neural network (CNN) based feature extractor is employed to extract distinctive local and global descriptors from the BEV images. Local descriptors are used to match BEV images with FAST keypoints for reprojection error construction, while global descriptors facilitate loop closure detection. Reprojection error minimization is then integrated with point-to-plane registration within an iterated Extended Kalman Filter (iEKF). In the back-end, global descriptors are used to create a KD-tree-indexed keyframe database for accurate loop closure detection. When a loop closure is detected, Random Sample Consensus (RANSAC) computes a coarse transform from BEV image matching, which serves as the initial estimate for Iterative Closest Point (ICP). The refined transform is subsequently incorporated into a factor graph along with odometry factors, improving the global consistency of localization. Extensive experiments conducted in various scenarios with different LiDAR types demonstrate that BEVLIO(LC) outperforms state-of-the-art methods, achieving competitive localization accuracy. Our code and video can be found at https://github.com/HxCa1/BEV-LIO-LC.

Recently, federated learning (FL) has been threat-ened by the gradient inversion attack that infers user-private data from shared gradients. To cope with this problem, the differential privacy (DP) technique is widely employed in FL. However, when FL faces the non-independent identically distributed (non-IID) data scenarios, applying DP to protect user data privacy remains inefficient in terms of model accuracy and communication costs. In this paper, inspired by the Mixup data augmentation method, we propose a privacy-preserving FL approach called Fed2VAEs to address this problem. Specifically, we introduce a Mixup Module consisting of two variational autoencoders to remove the private information of user data. To balance the trade-off between data privacy and data utility, from the perspective of mutual information, a learning objective is proposed. We conduct extensive experiments under different non-IID data settings, and the experimental results show that Fed2VAEs can significantly reduce the communication cost and improve model accuracy (up to 8.57%) on the premise of successfully protecting user data privacy.

The recent advancements in transformer-based visual trackers have led to significant progress, attributed to their strong modeling capabilities. However, as performance improves, running latency correspondingly increases, presenting a challenge for real-time robotics applications, especially on edge devices with computational constraints. In response to this, we introduce LiteTrack, an efficient transformer-based tracking model optimized for high-speed operations across various devices. It achieves a more favorable trade-off between accuracy and efficiency than the other lightweight trackers. The main innovations of LiteTrack encompass: 1) asynchronous feature extraction and interaction between the template and search region for better feature fushion and cutting redundant computation, and 2) pruning encoder layers from a heavy tracker to refine the balnace between performance and speed. As an example, our fastest variant, LiteTrack-B4, achieves 65.2% AO on the GOT-10k benchmark, surpassing all preceding efficient trackers, while running over 100 fps with ONNX on the Jetson Orin NX edge device. Moreover, our LiteTrack-B9 reaches competitive 72.2% AO on GOT-10k and 82.4% AUC on TrackingNet, and operates at 171 fps on an NVIDIA 2080Ti GPU. The code and demo materials will be available at https://github.com/TsingWei/LiteTrack.

The vehicle-road cooperative perception needs high accuracy and real-time automatic background filtering to separate background objects from foreground objects in complex traffic scenes. Reducing the influence of foreground objects to improve accuracy, and introducing a new framework to improve real-time performance are two main challenges in automatic background filtering. This paper proposes an Automatic Background Filtering method with innovative Frame Selection and Background Matrix Extraction modules (ABF-FSBME) to address these challenges. Firstly, a new space division method with equal hitting probability is proposed to divide the 3D point cloud formed by roadside Light Detection and Ranging (LiDAR), which can reduce the influence of slight LiDAR vibrations. Secondly, the terminal-edge-cloud framework is introduced to balance delay-constrained tasks and computation-intensive tasks in automatic background filtering. Thirdly, a variance-based frame selection strategy with a sliding window mechanism is proposed to select candidate frames with fewer foreground objects. This strategy can reduce the influence of foreground objects in a coarse-grained way. Meanwhile, a new background matrix extraction method is proposed to construct the background matrix. This method can further reduce the influence of foreground objects in a fine-grained way. Finally, based on the extracted background matrix from a cloud server, the edge server can filter the raw frame in real-time. The experimental results show that the proposed ABF-FSBME method has better accuracy than other methods in error rate and integrity rate. Besides, the proposed ABF-FSBME can complete fame filtering within 10ms, and has almost no network delay, so it can satisfy the real-time requirement.

Federated Learning (FL) has been widely used to facilitate distributed and privacy-preserving machine learning in recent years. Different from centralized training that usually has independent and identically distributed (IID) distribution of all users' data, FL suffers from significant communication cost and model performance degradation due to the non-IID data from individual edge devices. Existing work calibrates the local models using a global anchor or sharing global data. However, these studies either assume that the central server has the global dataset or require participating devices to share raw data, which incurs additional communication costs and privacy concerns. In this paper, we proposeSlaugFL, a novel selective GAN-based data augmentation scheme for communication-efficient edge FL, which selects representative devices to share specific local class prototypes with the central server for GAN model training and improves FL performance with the trained GAN. Specifically, on the server side, we generate diverse labeled candidate data with the help of powerful generative models (the stable diffusion model and ChatGPT). To ensure that the GAN-generated data possesses a similar domain to the devices' local data, we leverage these selected local class prototypes to pick desired GAN training samples from the labeled candidate data. On the device side, we propose a dual-calibration approach consisting of two calibration manners. Concretely, we augment devices' non-IID data with the trained GAN model, where devices utilize the trained GAN model to generate the IID dataset. Thus, the device's local model can be directly calibrated with the augmented data. With the generated IID data, we yield privacy-free (p-f) global class prototypes which can be employed to further calibrate devices' local models. Combining these two calibrations effectively improves devices' local models. Extensive experimental results show thatSlaugFLcan significantly reduce the communication cost (up to 52.49%) while achieving the same accuracy, compared to the state-of-the-art work.

Nowadays, face recognition technology has been dramatically boosted by the advances in deep learning and big data fields. However, this also poses grand challenges in protecting personal identity information in intelligent applications of the Internet of Things (IoT). Existing methods based on the$K$-Same algorithm have low effectiveness for protecting personal identity while preserving face attributes. In this article, we propose an attribute-preserving face de-identification framework called Enhanced Embedded AutoEncoders to address this problem. Our framework consists of three parts: 1) a privacy removal network (PRN); 2) a feature selection network; and 3) a privacy evaluation network. The main purpose of our framework is to ensure that the PRN is capable of discarding information involving identity privacy and retaining desired face attributes for certain prediction applications. In order to achieve this goal, the design of the PRN is crucial. Specifically, we employ two different autoencoders, one of which is embedded within the other. Extensive experimental results show that our framework outperforms existing methods by an average of 3.42%–26.22% in terms of data utility under comparable face de-identification performance, which indicates that the proposed framework can not only effectively retain face attributes but also protect personal identity well.

Edge computing and container technologies offer more possibilities for the development of Internet of Vehicles (IOV). However, many studies have neglected the dependencies among containers and the mobility of users. In this article, we propose a container migration strategy based on trajectory prediction, with the consideration of dependencies among containers. Given a set of containers with dependencies, we aim to reduce the service latency while distributing the containers as evenly as possible for load balance. Specifically, we leverage Recurrent Neural Network (RNN) to train a model for trajectory prediction. Based on the prediction, we can identify the location of the vehicles, after which we develop a mathematical model for service delay. We formulate the problem as a 0-1 program and solve the problem by Hunger Games Search (HGS). The proposed algorithm is validated with a real vehicle dataset and an Alibaba cluster dataset. Experiment results demonstrate that we can predict the trajectory of vehicle accurately and the container migration strategy can effectively reduce service latency and improve server load balancing, compared to alternative algorithms. In addition, we characterize the impact of error correction mechanism and also the effect of bandwidth on the optimization.

The distribution of large-size data block in the Internet of Vehicles (IoV), especially in the urban IoV with dense vehicles, is still a challenge issue. Though the methods based on 5G-cellular network commonly can efficiently distribute the large-size file in IoVs, they also have obvious defects, such as occupying scarce 5G resources, generating communication fees and limited service coverage. This paper focuses on distributing large-size data block in IoVs based on dedicated vehicular ad hoc network to reduce the relying on cellular (5G/6G) resources. A heterogeneous vehicular network (HetVNET), in which a short-range OFDM wideband communication (SOWC) with ultra-high data rate and the original IEEE 802.11p protocol are included, is first proposed. To match with the proposed HetVNET, a content-centric data distribution scheme based on edge caching is designed. The content-centric data distribution architecture for distributing large-size file, the cache node selection and the process of data distribution and control based on the HetVNET are studied. The 5G/6G communication can be enabled in the scheme to further enhance the engineering stability of the massive infrastructure-to-vehicle (I2V) broadcast in IoVs. The evaluation results show that the proposed approach has low delivery delay and high penetration ratio.

Summary Selective maintenance is often applied in many industrial environments where maintenance is performed between sequence missions. When the mission time is stochastic and there are multiple maintenance workers with different capacities, the system reliability of the next work mission can be maximized by using a stochastic model under the constraint of the limit maintenance time. The optimal maintenance strategy is obtained with an optimization algorithm. A simulation was performed to verify the validity and feasibility of the proposed model.