Indoor Visible Light Positioning Based on Extreme Learning Machine with Local Training Samples

Multi-directional light-emitting diode aided visible light positioning system exploiting regular pyramid: design and demonstration

“…8(b), we can see that the proposed MD-nVLP system exhibits a further deterioration in LE performance, resulting on an average LE of 19.1 cm, this phenomenon further confirms that diffuse reflection may be the primary factor contributing to the degradation of LE performance. In the next work, we intend to solve this issue through the utilization of ML 22 , 23 and ANN 24 – 27 techniques.…”

“…The proposed RSSR based LLS algorithm is concise, performs in real-time and avoids the excessive offline data collection campaign compared to those ML/ANN based algorithms 22 – 27 …”

“…In this case, the traditional radio frequency based localization algorithms can be employed into achieving accurate positioning results, such as received signal strength (RSS), 10 – 13 received signal strength ratio (RSSR), 14 , 15 time-of-arrival, 16 time-difference-of-arrival (TDOA), 17 , 18 angle-of-arrival, 19 , 20 and fingerprint scheme 21 . Furthermore, to mitigate the adverse effects of inaccurate model and light intensity variation on the system performance, classical intelligent algorithms are increasingly utilized by researchers to solve the localization problem of the nVLP model, such as machine learning (ML), 22 , 23 artificial neural network (ANN), 24 – 27 and some heuristic algorithms 28 – 35 Specifically, in Ref.…”

“…26, the gated recurrent unit neural network (GRU-NN) was employed to deal with the nonlinearity of the nVLP system, resulting in an improvement of average localization accuracy. As a type of ANN, the extreme learning machine (ELM) can immediately complete the parameter training to achieve satisfactory performance, therefore, it is increasingly utilized in nVLP systems 23 , 27 , 33 . In Ref.…”

“…As a type of ANN, the extreme learning machine (ELM) can immediately complete the parameter training to achieve satisfactory performance, therefore, it is increasingly utilized in nVLP systems. 23,27,33 In Ref. 33, two heuristic algorithms of grey wolf optimizer (GWO) and particle swarm optimization (PSO) were adopted to optimize the ELM parameters respectively in order to obtain higher classification accuracy, namely, EML-GWO and EML-PSO algorithms, which have a better positioning performance and robustness than the basic EML algorithm.…”