Acta Optica Sinica, Volume. 44, Issue 22, 2206004(2024)

Multi-Directional Light-Emitting Diodes Aided Visible Light Positioning Scheme Based on Neural Network

Qingxiang Li1, Haiyan Liu1, Yihui Chen1, Yi Cheng1, Qianqian Luo1、*, and Zhengpeng Li1,2、**
Author Affiliations
  • 1School of Physics and Electronic Engineering, Hubei University of Arts and Science, Xiangyang441053, Hubei , China
  • 2Hubei Provincial Engineering Research Center of Emergency Communication Technology and System, Xiangyang441053, Hubei , China
  • show less
    Figures & Tables(14)
    A typical example of MD-LED integrated lamp utilizing pyramid structure and four planar light sources
    Proposed ANN model based on RSSR
    Flow chart of RSSR-based ANN model
    Normalized light intensity radiation curves of different types of LEDs. (a) Standard Lambert radiation LED; (b) XLamp®XR-E LED; (c) Z Power LED
    Simulated LE performance of different types of LEDs adopting RSSR-ANN model (α and γ are tilted angles of PD and LED, respectively). (a)(b) Adopt the LED shown in Fig. 4(a); (c)(d) utilize the LED shown in Fig. 4(b); (e)(f) employ the LED illustrated in Fig. 4(c)
    Two prototypes of MD-LED lamps for experimental verification. (a) Utilizing commercially available Lambert radiation LED chips; (b) employing commercially available LED small downlights
    Experimental scenario[20]
    Measured LE performance of proposed scheme exploiting the integrated lamp in Fig. 6. (a)(b) Adopt the MD-LED integrated lamp shown in Fig. 6(a); (c)(d) utilize the MD-LED integrated lamp shown in Fig. 6(b)
    LE performance measured in the reflection scenario. (a) Reflection scenario[20]; (b) measured performance
    • Table 1. Neural network model parameters used in simulation

      View table

      Table 1. Neural network model parameters used in simulation

      ParameterSetting value
      Nodes number of input layer6
      Nodes number of hidden layer 112
      Nodes number of hidden layer 230
      Nodes number of hidden layer 32
      Nodes number of output layer2
      Transfer functiontanh
      Maximum iteration number5000
      Learning rate10-3
      Training dataset81
      Testing dataset441
      ANN optimizerPytorch Adam Optimizer
    • Table 2. Main simulation parameters

      View table

      Table 2. Main simulation parameters

      ParameterSetting value
      Room size /(m×m×m )4×4×3.2
      Number of LEDs N4
      Integrated lamp size hL / cm12
      LED optical power PL / W2
      Lambertian radiation lobe mode number n1
      LED tilt angle γ /(°)15‒30
      Horizontal distance between LEDs rh /cm1.33
      PD receiving area AU /cm21
      PD responsivity R /(A·W-10.4
      PD polar angle α /(°)0‒90
      PD azimuth angle β /(°)0‒360
      Noise power or variance σ2 /A210-13 [11
      PD height above the floor hU /m1.14
    • Table 3. Neural network model parameters employed in experiment

      View table

      Table 3. Neural network model parameters employed in experiment

      ParameterSetting value
      Nodes number of input layer6 or 10
      Nodes number of hidden layer 110
      Nodes number of hidden layer 214
      Nodes number of hidden layer 32
      Nodes number of output layer2
      Transfer functiontanh
      Maximum iteration number5000
      Learning rate10-3
      Training dataset121
      Testing dataset441
      ANN optimizerPytorch Adam Optimizer
    • Table 4. Average LE comparison between RSSR-ANN and RSSR-LLS

      View table

      Table 4. Average LE comparison between RSSR-ANN and RSSR-LLS

      ScenarioPD stateAverage LE /cmReduction rate /%
      RSSR-LLS20RSSR-ANN
      Least reflectionHorizontal10.72.9672.3
      Least reflectionRandomly tilted11.65.5152.5
      Random reflectionHorizontal19.16.0268.5
    • Table 5. Comparison among the existing indoor VLP technologies

      View table

      Table 5. Comparison among the existing indoor VLP technologies

      Ref.SchemeAverage LE /cmCoverage area /(m×m×m)Scheme and device complexitySupport receiver tilt or not
      This workRSSR-ANN3‒64×4×3.2MediumYes
      20RSSR-LLS11‒194×4×3.2LowYes
      13GA-ELM0.924×4×3MediumNo
      14ISSA-ELM1.0‒3.95×5×3HighNo
      15Elman-ANN7.133.6×3.6×3HighNo
      16GA-CNN4.114×4×2.5HighNo
      17BAS-ANN<40.8×0.8×0.8HighNo
      18ELM1.174×4×3MediumNo
    Tools

    Get Citation

    Copy Citation Text

    Qingxiang Li, Haiyan Liu, Yihui Chen, Yi Cheng, Qianqian Luo, Zhengpeng Li. Multi-Directional Light-Emitting Diodes Aided Visible Light Positioning Scheme Based on Neural Network[J]. Acta Optica Sinica, 2024, 44(22): 2206004

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Fiber Optics and Optical Communications

    Received: Jun. 5, 2024

    Accepted: Aug. 1, 2024

    Published Online: Nov. 20, 2024

    The Author Email: Qianqian Luo (qqluo@hbuas.com), Zhengpeng Li (zpli@hbuas.com)

    DOI:10.3788/AOS241136

    CSTR:32393.14.AOS241136

    Topics