Industrial time series analysis is the core foundation of equipment status monitoring and industrial intelligence. However, the long-tailed distribution characteristics of time series caused by low-frequency and low-probability events seriously restrict the performance of analysis models. In key scenarios such as industrial process prediction and anomaly detection, data analysis models face obvious performance bottlenecks due to insufficient representation of tail events. Existing data augmentation methods have dual limitations in capturing tail patterns and modeling long-distance time series dependencies. To address this challenge, this paper proposes an industrial long-tailed time series generator (DLTTS) model based on a diffusion model. Firstly, a hybrid architecture generation model is constructed to deeply integrate the encoder-decoder informer structure with the traditional diffusion process to maintain long-distance time patterns and achieve long-distance time series output; Secondly, we propose Fourier-based batch-Monte Carlo (FBMC) loss to enhance the model's ability to capture low-frequency events, thereby improve the quality of tail time series generation. Experiments show that DLTTS maintains the authenticity and diversity of tail time series generation in industrial-grade long-tail time series generation tasks. It also exhibits robust performance in imputation and forecasting tasks, verifying the multiple performance advantages of this method for cross-task application.

Mobile robots necessitate advanced natural language understanding capabilities to accurately identify locations and perform tasks such as package delivery. However, traditional visual place recognition (VPR) methods rely solely on single-view visual information and cannot interpret human language descriptions. To overcome this challenge, we bridge text and vision by proposing a multiview (360° views of the surroundings) text-vision registration approach called Text4VPR for place recognition task, which is the first method that exclusively utilizes textual descriptions to match a database of images. Text4VPR employs the frozen T5 language model to extract global textual embeddings. Additionally, it utilizes the Sinkhorn algorithm with temperature coefficient to assign local tokens to their respective clusters, thereby aggregating visual descriptors from images. During the training stage, Text4VPR emphasizes the alignment between individual text-image pairs for precise textual description. In the inference stage, Text4VPR uses the Cascaded Cross-Attention Cosine Alignment (CCCA) to address the internal mismatch between text and image groups. Subsequently, Text4VPR performs precisely place match based on the descriptions of text-image groups. On Street360Loc, the first text to image VPR dataset we created, Text4VPR builds a robust baseline, achieving a leading top-1 accuracy of 56% and a leading top-10 accuracy of 91% within a 5-meter radius on the test set, which indicates that localization from textual descriptions to images is not only feasible but also holds significant potential for further advancement, as shown in Figure 1. Our code is available at https://github.com/nuozimiaowu/Text4VPR.

In recent years, with the rapid development of automated driving technology, the safety and quality of automated driving software have become important concerns.Prior studies have shown that code clones are prevalent in autonomous driving software and significantly impact software quality, maintenance costs, and module independence.Although code cloning can improve development productivity, they are more likely to introduce co-modifications and software defects, which leads to high detection and fixing costs.In this paper, we propose a novel approach to predict defective code clones in autonomous driving software by focusing specifically on clone code snippets as prediction targets.We designed a comprehensive set of nine code metrics characterizing clone snippets from three dimensions: code size, complexity, and historical modifications.Using six machine learning algorithms, we built prediction models and evaluated them on two representative L4 opensource autonomous driving platforms: Apollo and Autoware.Our experimental results demonstrate that the developed prediction models achieve good performance with accuracy rates ranging from 82.7% to 94.4%.Through principal component analysis, we identified that the number of added lines, modification frequency, and clone line count are the most significant factors contributing to defect proneness in code clones.We believe that this work enables developers to locate defective code clones more precisely.Thus, they can improve testing efficiency, reduce bug fix costs, and ultimately improve the quality and reliability of autonomous driving software.

Task assignment and path planning are crucial links in the task execution of uncrewed aerial vehicle (UAV) cluster, especially in high-dimensional complex scenarios, the calculation difficulty increases significantly. To solve this problem, swarm intelligence as an efficient strategy emerged. In order to solve the challenge of incomplete information in the task assignment of UAVs and the problems of intra-group cooperation and competition, we propose an innovative multi-agent near-end strategy optimization algorithm (MAPPO algorithm). The algorithm is designed for the task assignment of UAV in incomplete information environment. By constructing a practical algorithm model and combining the incomplete information game theory, the original algorithm is optimized to better deal with the cooperation and competition mechanism within the UAVs. Secondly, the global search capability is poor and local optimization is easy to occur. The dynamic cluster particle swarm optimization (DCPSO) algorithm is proposed to model the task scenario of UAVs path planning problem by using artificial potential field method and rolling time domain control principle. Tent chaos mapping and dynamic cluster mechanism are introduced to further improve the global search capability and search accuracy. Finally, DCPSO algorithm is used to optimize the objective function of the model, and the selection of UAV trajectory points is obtained. Simulation results under different combinations of single-peak/multi-peak, low-dimensional/high-dimensional benchmark test functions show that DCPSO algorithm has better optimization ability, mean value and variance compared with PSO, pigeon inspired optimization (PIO), Sparrow search algorithm (SSA) and chaotic disturbed pigeon flock optimization (CDPIO) algorithms. Better search accuracy and stability.

In this paper, a starry image matching method based on isomorphism sub graph and LCS (Longest Common Sub-sequence) is proposed for dim and small space target detecting. A starry image mainly consists of a background with a large number of low-gray pixels and bright but small facular. The relative location of the stars can be regarded as a feature set for matching in the form of a graph. While matching two graphs, the LCS is applied for measuring the similarity and acquiring the isomorphism sub graph. Due to the characteristic of the starry image, this method could deal with the images with more rotation and shift. Experimental results show that this method can provide a good way for the trajectory acquisition of dim and small targets.