Xinheng Wang’s research while affiliated with Xi’an Jiaotong-Liverpool University and other places

Viewpoint planning determines the accuracy, processing speed, and lightweight of structure from motion. Despite the importance of viewpoint planning optimization to industrial digital services, existing methods show evident shortages in balancing between the reconstruction accuracy and the viewpoint number. Hence, this paper defines a new next-best-view problem for structure from motion, which aims to improve the accuracy, reduce the viewpoint number, and strike a balance between the two, simultaneously. Besides, to resolve the problem, this paper presents a novel viewpoint planning optimization method based on Proximal Policy Optimization. This method incorporates double models, action mask, and sim-to-real training to improve the training efficiency. Additionally, this method applies transfer-learning and fine-tuning to improve the versatility of the optimized viewpoint plan. A case study and experiments with multiple house models illustrate the method. In the experiment, the optimized viewpoint plan achieved 12.42%, 14.87%, 16.39%, 15.58%, and 32.35% reduction in Chamfer Distance, Earth Mover’s Distance, the viewpoint number, the file size, and reconstruction processing time compared to the naïve baseline, respectively. Also, compared to existing methods, the proposed method showed advantages from different perspectives, particularly in the balance between the reconstruction accuracy and the viewpoint number.

Accurate vibration measurement is crucial for monitoring and diagnosing industrial equipment. Existing solutions require either installing contact sensors on the equipment or using non-contact sensors such as laser. Both approaches involve complex deployment, stringent environmental conditions, and high cost. As a better alternative, we propose a sub-millimeter acoustic vibration measurement system using a single smartphone, called Mobile-Vib. Firstly, we develop a novel acoustic ranging method that builds on traditional acoustic ranging techniques, incorporating the reflection principle of acoustic signals from vibrating objects. This approach addresses the challenge of acoustic signal refresh rate in vibration measurement by employing advanced signal design and processing techniques. Secondly, we design a noise removal algorithm utilizing the dual-channel technology of smartphones to minimize multipath signals and noise interference, enabling accurate phase estimation. To mitigate the impact of unrelated human motions in real-world measurements, we implement an optimisation-based method to correct distortions and reduce errors. Finally, by clarifying the relationship between phase changes and actual displacement, we enable tracking of vibration displacement in industrial environments. We have implemented Mobile-Vib, and the extensive experimental results demonstrate an average error of approximately 0.629 mm in displacement estimation and 5.6 Hz in frequency estimation at a 1-meter distance from the vibrating object in real industrial vibration monitoring scenarios.

As communication and computing technologies advance, vehicular edge computing emerges as a promising paradigm for delivering a wide array of intelligent services in 6G enabled Intelligent Autonomous Transport Systems. These service requests, are safety-oriented and typically require the fusion of processing results from multiple independent computation tasks generated by various onboard sensors, in which the computation tasks are delay-sensitive and computation-intensive. Consequently, the allocation of multiple tasks within a single service request while efficiently reducing request completion time and energy consumption presents a substantial challenge. In order to address the problem of multi-task simultaneous scheduling, this paper proposed to employ deep reinforcement learning and edge computing architecture to make task scheduling decisions for vehicles. Firstly, the Vehicle-Infrastructure Network (VINET) is designed, in which the vehicles can assign multiple tasks to the edge servers and other idle vehicles, thus extending the task processing capabilities for vehicles. Secondly, Fully-decentralized Multi-agent Proximal Policy Optimization (FMPPO) algorithm is proposed to make task scheduling decisions for autonomous driving, the large model trained via FMPPO is adaptable to different scenarios with various numbers of vehicles. Thirdly, by taking into account task characteristic, environmental status, and vehicle mobility, the proposed method can make task scheduling decisions in real-time and then dynamically distributes tasks based on the decisions. Finally, experimental results demonstrate that the designed method outperforms benchmark methods in terms of both completion time and energy consumption of computation tasks.

With the advancement of edge computing, more and more intelligent applications are being deployed at the edge in proximity to end devices to provide in-vehicle services. However, the implementation of some complex services requires the collaboration of multiple AI models to handle and analyze various types of sensory data. In this context, the simultaneous scheduling and execution of multiple model inference tasks is an emerging scenario and faces many challenges. One of the major challenges is to reduce the completion time of time-sensitive services. In order to solve this problem, a multiagent reinforcement learning-based multimodel inference task scheduling method was proposed in this article, with a newly designed reward function to jointly optimize the overall running time and load imbalance. First, the multiagent proximal policy optimization algorithm is utilized for designing the task scheduling method. Second, the designed method can generate near-optimal task scheduling decisions and then dynamically allocate inference tasks to different edge applications based on their status and task characteristics. Third, one assessment index, quality of method, is defined and the proposed method is compared with the other five benchmark methods. Experimental results reveal that the proposed method can reduce the running time of multimodel inference by at least 25% or more, closing to the optimal solution.

Near-ultrasound acoustic localization has emerged as a cost-effective and precise technique for indoor localization. Acoustic Non-Line-of-Sight (NLOS) identification is essential in indoor localization systems. Current NLOS identification approaches predominantly utilize Matched Filter (MF) characteristics or signal spectrum output. However, these methods often fail to deliver adequate identification performance in cross- scenario environments, including those that employ statistical analysis and Deep Learning (DL) techniques. In this paper, we propose a novel NLOS identification method specifically designed for near-ultrasound localization signals. Our approach utilizes an enhanced Frequency-Modulated Continuous Wave (FMCW) technique to obtain an improved intermediate frequency (IF) signal. Subsequently, a subset of eight distinct features is extracted and selected from the IF signal. The features are then employed to enhance an XGBoost classifier for the identification of NLOS conditions. Experimental results across three scenarios demonstrate that our method attains a classification accuracy of 99.9% within identical-room settings, and achieves accuracies of 93.73% and 86.20% in cross-scenario environments, respectively. Additionally, the extracted features demonstrate promising performance not only within the presented XGBoost classifier but also across various statistical machine-learning models.

High-accuracy Time of Arrival (TOA) estimation is essential for indoor acoustic localization, but traditional TOA estimation methods often fail to obtain accurate and robust TOA estimation performance due to severe multipath challenges such as peak extraction by General Cross Correlation (GCC). To address the problem, this paper presents a novel acoustic TOA estimation technique called FDME-RATE, a high-accuracy acoustic ranging system in acoustic localization. The proposed method utilizes GCC to determine the coarse signal arrival time, followed by an enhanced Frequency Modulated Continuous Wave (FMCW) technique to convert the various arrival paths into distinct frequencies within the spectrum. Ultimately, we employ an Adaptive Multiple Signal Classification (AMUSIC) algorithm to achieve high-resolution TOA estimation. Through simulations and experiments conducted in a reverberant corridor, the results indicate that our method achieves an average TOA ranging error of just 0.16 m and 0.19 m at the 95% confidence level across a 30-m span. Within a 20-m range, there is only a centimeter-level error. Additionally, in non-line-of-sight (NLOS) scenarios, our proposed method further enhances TOA-based ranging accuracy by an additional 0.3 m over the same distance. The experimental results also demonstrate that our method outperforms traditional methods and recent research in terms of both estimation accuracy and robustness.

Optimizing anchor deployment is critical to ensure the performance and positioning stability of indoor positioning systems in real-world applications. In this paper, a new anchor deployment optimization method is proposed to enhance the positioning performance of range-based positioning systems in the Non-Line-of-Sight (NLOS) environment without increasing the application cost. Firstly a new fitness function is proposed by simultaneously considering the mean geometric dilution of precision (GDOP) and the coverage of available positioning area in the indoor NLOS environment. Then, a search architecture based on a particle swarm optimization (PSO) algorithm is proposed to optimize anchor deployment. The initialization method of swarm’s position and velocity is given, and the calculation process of the search architecture is introduced in detail. The results obtained from numerical simulations and experimental investigations verified that, for range-based positioning systems in NLOS environment, the accuracy and stability can be significantly improved by optimizing the anchor deployment through our proposed method.