稀疏训练指纹库融合MMPSO-ELM室内可见光定位

In order to solve the shortcomings of indoor visible light positioning method using extreme learning machine (ELM) neural network, such as large errors, long training time of network model and poor stability of results, a sparse training fingerprint database was adopted, multi-objective momentum particle swarm optimization (MMPSO) was integrated, and the indoor visible light positioning method of ELM was combined to form an MMPSO-ELM scheme. The momentum factor is introduced to avoid excessive oscillation during iteration and speed up the system convergence. The training data sets are randomly selected in different positioning Spaces, and the proposed scheme is compared with three positioning algorithms, back-propagation (BP), ELM and PSO-ELM, under the condition of different number of test points. The results show that the MMPSO-ELM scheme can predict the positioning of 80 groups of undetermined sites under the condition of 20 groups of training data. The maximum positioning error is 0.0225m, the minimum error is 0.00093m, and the average positioning error is as low as 0.00143m. Moreover, the positioning performance is little affected by the positioning space size. The MMPSO-ELM visible light positioning scheme has the advantages of high positioning accuracy, fast speed and strong generalization. This study provides theoretical support for realizing fast and accurate location in indoor places.