Scholarlink

<p>To address the challenges of high fingerprint database construction costs and insufficient accuracy of single model approaches in indoor visible light positioning (VLP) systems, this paper proposes a novel positioning algorithm based on a sparse fingerprint database, integrating extreme learning machine (ELM) and support vector regression (SVR) with error confidence fusion. By generating a high density virtual fingerprint database through sparse physical sampling combined with interpolation techniques, the offline data collection workload is significantly reduced. Leveraging the efficient training capability of ELM and the strong generalization performance of SVR, a dynamic error confidence-based weighting strategy is designed to adaptively balance the contributions of the two models, thereby improving positioning accuracy and robustness in complex environments. Simulation results demonstrate that, in a 3 × 4 × 3&#xa0;m indoor scenario, the proposed algorithm achieves a mean positioning error of 4.2&#xa0;cm and maintains stable accuracy under low signal to noise ratio (SNR) conditions. Compared with existing methods, the proposed approach exhibits significant advantages in both accuracy and robustness, offering a feasible solution for low cost, high precision indoor VLP systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

ELM-SVR error confidence fusion algorithm based on sparse fingerprint database for indoor visible light positioning system

  • 期刊论文

  • 发表日期：

    2025年7月17日

查看全文
  • Lieping Zhang,
  • Yao Chen,
  • Huaquan Gan,
  • Shenpeng Huang

摘要

To address the challenges of high fingerprint database construction costs and insufficient accuracy of single model approaches in indoor visible light positioning (VLP) systems, this paper proposes a novel positioning algorithm based on a sparse fingerprint database, integrating extreme learning machine (ELM) and support vector regression (SVR) with error confidence fusion. By generating a high density virtual fingerprint database through sparse physical sampling combined with interpolation techniques, the offline data collection workload is significantly reduced. Leveraging the efficient training capability of ELM and the strong generalization performance of SVR, a dynamic error confidence-based weighting strategy is designed to adaptively balance the contributions of the two models, thereby improving positioning accuracy and robustness in complex environments. Simulation results demonstrate that, in a 3 × 4 × 3 m indoor scenario, the proposed algorithm achieves a mean positioning error of 4.2 cm and maintains stable accuracy under low signal to noise ratio (SNR) conditions. Compared with existing methods, the proposed approach exhibits significant advantages in both accuracy and robustness, offering a feasible solution for low cost, high precision indoor VLP systems.