Dynamical Optimization Of Lighting Layout For Indoor Visible Light Communication Based On Deep Learning

In recent years,visible light communication has become popular in the field of optical communication due to its wide bandwidth,high security and non-electromagnetic radiation.In an indoor visible light communication system,the lighting infrastructure design should not be fixed in single situation.It should be dynamical because of the time-varying environment changes.For examples,the movement of obstructions or the ambient light noise generated by windows will bring extra dead zone in communication.They will also fluctuate the distribution of the signal-to-noise ratio?SNR?.The dynamical design of lighting infrastructure,which focus on improving SNR in a changeable room,has been studied in this paper.By the effort of the genetic algorithm,it can reduce the dead zone by optimizing lamp distribution in any room status.In other words,the effective areas can achive the maximum.The optimized lamps layout also can be found.In low data rate,like10Mb/s,effective areas which bit error rate?BER?below to10^-6are between94.85%to97.28%in the rooms with obstructions.Or the effective area can achieve91.04%in the room with ambient light.However,when the data rate is higher,like100Mb/s,the effective areas are between77.92%to90.9%.Besides that,the neural network also been used in predict the layout.The multi-layer perception?MLP?and convolutional neural networks?CNN?were used in training process for speed up the calculations compare to the genetic algorithm.After a series of steps of training the model accuracy can achieve85%and calculate the distribution in 2 milliseconds,thus it is capable of dynamical optimized the lighting infrastructure design,rather than wait long time for the result which generate by genetic algorithm.