{"title":"用于室内惯性导航的记忆管理循环Kolmogorov-Arnold网络","authors":"Qiaolin Pu;Yunhai Li;Mu Zhou","doi":"10.1109/TIM.2025.3604962","DOIUrl":null,"url":null,"abstract":"Although inertial measurement units (IMUs) have emerged as a promising solution for indoor positioning, due to their low cost, energy efficiency, and infrastructure-independent nature, the accumulation of measurement errors remains a critical challenge that hinders their widespread application. In recent years, with the advancement of deep learning, data-driven methods have been proven to effectively solve this issue through trained neural networks. However, current mainstream data-driven models predominantly adopt simple multilayer perceptron (MLP) architectures. These architectures exhibit inherent limitations in scalability and interpretability when processing IMU data, resulting in inefficient parameter utilization and suboptimal positioning accuracy in inertial navigation applications. Hence, this article proposes a novel data-driven model named KANet, which consists of two key components: recurrent Kolmogorov–Arnold networks (RKANs) and long short-term memory (LSTM)-inspired memory management units. RKAN fundamentally integrates the powerful function approximation capability of Kolmogorov–Arnold networks (KANs) with the sequential modeling strength of recurrent neural networks (RNNs), while the LSTM-inspired memory mechanisms enhance temporal dependency modeling in long sequences. By leveraging learnable activation functions for nonlinear pattern representation with effective memory retention characteristics, this architecture enhances the model’s capacity to process complex sequential IMU data. This innovative design overcomes the limitations of traditional models in processing complex sequence patterns, demonstrating competitive accuracy and improved parameter efficiency in multistep time-series prediction. Experimental results demonstrate that the proposed framework achieves superior performance compared with traditional PDR models and state-of-the-art neural networks, exhibiting higher parameter efficiency across diverse datasets while maintaining high positioning accuracy. In addition, comprehensive benchmarking against tactical-grade IMU systems reveals that KANet-assisted consumer-grade MEMS sensors can achieve positioning accuracy that, in certain aspects, approaches or even matches the performance levels of higher precision IMU systems.","PeriodicalId":13341,"journal":{"name":"IEEE Transactions on Instrumentation and Measurement","volume":"74 ","pages":"1-13"},"PeriodicalIF":5.9000,"publicationDate":"2025-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"KANet: Memory-Managed Recurrent Kolmogorov–Arnold Network for Indoor Inertial Navigation\",\"authors\":\"Qiaolin Pu;Yunhai Li;Mu Zhou\",\"doi\":\"10.1109/TIM.2025.3604962\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Although inertial measurement units (IMUs) have emerged as a promising solution for indoor positioning, due to their low cost, energy efficiency, and infrastructure-independent nature, the accumulation of measurement errors remains a critical challenge that hinders their widespread application. In recent years, with the advancement of deep learning, data-driven methods have been proven to effectively solve this issue through trained neural networks. However, current mainstream data-driven models predominantly adopt simple multilayer perceptron (MLP) architectures. These architectures exhibit inherent limitations in scalability and interpretability when processing IMU data, resulting in inefficient parameter utilization and suboptimal positioning accuracy in inertial navigation applications. Hence, this article proposes a novel data-driven model named KANet, which consists of two key components: recurrent Kolmogorov–Arnold networks (RKANs) and long short-term memory (LSTM)-inspired memory management units. RKAN fundamentally integrates the powerful function approximation capability of Kolmogorov–Arnold networks (KANs) with the sequential modeling strength of recurrent neural networks (RNNs), while the LSTM-inspired memory mechanisms enhance temporal dependency modeling in long sequences. By leveraging learnable activation functions for nonlinear pattern representation with effective memory retention characteristics, this architecture enhances the model’s capacity to process complex sequential IMU data. This innovative design overcomes the limitations of traditional models in processing complex sequence patterns, demonstrating competitive accuracy and improved parameter efficiency in multistep time-series prediction. Experimental results demonstrate that the proposed framework achieves superior performance compared with traditional PDR models and state-of-the-art neural networks, exhibiting higher parameter efficiency across diverse datasets while maintaining high positioning accuracy. In addition, comprehensive benchmarking against tactical-grade IMU systems reveals that KANet-assisted consumer-grade MEMS sensors can achieve positioning accuracy that, in certain aspects, approaches or even matches the performance levels of higher precision IMU systems.\",\"PeriodicalId\":13341,\"journal\":{\"name\":\"IEEE Transactions on Instrumentation and Measurement\",\"volume\":\"74 \",\"pages\":\"1-13\"},\"PeriodicalIF\":5.9000,\"publicationDate\":\"2025-09-02\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"IEEE Transactions on Instrumentation and Measurement\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://ieeexplore.ieee.org/document/11146884/\",\"RegionNum\":2,\"RegionCategory\":\"工程技术\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"ENGINEERING, ELECTRICAL & ELECTRONIC\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Instrumentation and Measurement","FirstCategoryId":"5","ListUrlMain":"https://ieeexplore.ieee.org/document/11146884/","RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
KANet: Memory-Managed Recurrent Kolmogorov–Arnold Network for Indoor Inertial Navigation
Although inertial measurement units (IMUs) have emerged as a promising solution for indoor positioning, due to their low cost, energy efficiency, and infrastructure-independent nature, the accumulation of measurement errors remains a critical challenge that hinders their widespread application. In recent years, with the advancement of deep learning, data-driven methods have been proven to effectively solve this issue through trained neural networks. However, current mainstream data-driven models predominantly adopt simple multilayer perceptron (MLP) architectures. These architectures exhibit inherent limitations in scalability and interpretability when processing IMU data, resulting in inefficient parameter utilization and suboptimal positioning accuracy in inertial navigation applications. Hence, this article proposes a novel data-driven model named KANet, which consists of two key components: recurrent Kolmogorov–Arnold networks (RKANs) and long short-term memory (LSTM)-inspired memory management units. RKAN fundamentally integrates the powerful function approximation capability of Kolmogorov–Arnold networks (KANs) with the sequential modeling strength of recurrent neural networks (RNNs), while the LSTM-inspired memory mechanisms enhance temporal dependency modeling in long sequences. By leveraging learnable activation functions for nonlinear pattern representation with effective memory retention characteristics, this architecture enhances the model’s capacity to process complex sequential IMU data. This innovative design overcomes the limitations of traditional models in processing complex sequence patterns, demonstrating competitive accuracy and improved parameter efficiency in multistep time-series prediction. Experimental results demonstrate that the proposed framework achieves superior performance compared with traditional PDR models and state-of-the-art neural networks, exhibiting higher parameter efficiency across diverse datasets while maintaining high positioning accuracy. In addition, comprehensive benchmarking against tactical-grade IMU systems reveals that KANet-assisted consumer-grade MEMS sensors can achieve positioning accuracy that, in certain aspects, approaches or even matches the performance levels of higher precision IMU systems.
期刊介绍:
Papers are sought that address innovative solutions to the development and use of electrical and electronic instruments and equipment to measure, monitor and/or record physical phenomena for the purpose of advancing measurement science, methods, functionality and applications. The scope of these papers may encompass: (1) theory, methodology, and practice of measurement; (2) design, development and evaluation of instrumentation and measurement systems and components used in generating, acquiring, conditioning and processing signals; (3) analysis, representation, display, and preservation of the information obtained from a set of measurements; and (4) scientific and technical support to establishment and maintenance of technical standards in the field of Instrumentation and Measurement.