A Study Of Transmission Technology For Multi-color Multi-user Visible Light Communication Systems

The optimal design of visible light communication(VLC)systems should take both illumination quality and communication performance into considerations.When considering both of them,multi-color(MC)LED has advantages on adjustable correlated color temperature and high-rate transmittion.Considering using MC LED as the optical sources,only end-to-end communication scenarios are considered in most existing research works.For multi-user(MU)VLC scenarios,the studies of the transmission technologies are complex,which should consider both the MU communication performance criterias and the constraints of the white light by mixing red,green and blue emitters.To solve the problem,in this thesis,the study of the transmission technology is conducted from different angles,i.e.,linear precoding design,linear transceiver design,user-grouping scheme design,constellation design and non-linear transceiver design.The main content and contribution of this paper are presented as following.1.Zero-forcing(ZF)based linear precoding design.To suppress the impact of multi-user interference(MUI)and multi-color interference(MCI)on the system performance,a ZF precoding design is proposed based on the model of the MC-MU VLC downlink broadcast channels.The target of the precoding design is obtaining the achievable data rate with the considerations on two typical performance criteria in MU communication systems,i.e.,the max-min fairness and the maximum sum-rate.The establishment of the optimization problem is based on the upper and the lower bounds of the channel capacity in the amplitude-limited scalar additive Gaussian channel.Considering the two criteria and two kinds of bounds,there are four optimization problems in total.To solve the four non-convex optimization problems,we cast the non-convex optimization problems into convex ones with the help of slack variables and Taylor expansions.Correspondingly,iterative algorithms are developed.2.Mean-squared error(MSE)based linear transceiver design.To improve the system performance further,we consider MUI,MCI and the additive noise at receiver together and propose a joint design of the precoding matrix at transmitter and the decorrelator matrix at receiver.Considering the user fairness criterion,the optimization of the transceiver design is estabalished with the target of minimizing the maximum mean-squared error(MSE)under the constraints on color,illumination flux and signal amplitude.To solve the problem,an iterative algorithm is proposed.In each iteration of the algorithm,only simple matrix operation and convex program are needed.Simulation results show that the proposed transceiver design owns a better performance than the ZF precoding shceme.3.User-grouping algorithm based on a linear transceiver design.Considering the case that a severe decrease of system performance is caused by the MUI and using only precoding method can't meet the requirement of symbol error rate,we propose a two-stage tree-based usergrouping algorithm with usage of multi-color resources,which comes to a trade-off between transmission efficiency and system rubust.In the first stage,spatial correlation between users is taken as the criterion and a tree-based user-grouping algorithm is proposed.In the second stage,multi-color resources are allocated to different groups according to the sizes of the groups and the mixing ratio of multi-color lights.Based on the results of user-grouping method,a transceiver design is proposed which makes the users in a group be able to access to the network simultaneously,4.Multi-user joint constellation(JC)design for color-shift keying(CSK).With the user fairness consideration,a multi-user JC design for CSK is proposed with the target of maximizing the minimum Euclidean distance(MED)between different receiving constellation points.In the JC design,the generation of transmitted symbols are joint in our design while independent in traditional multiple-input multiple-output(MIMO)systems.To solve the non-convex problem,an iterative algorithm with the help of slack variables and Taylor expansions is proposed,which provides a method to approach the solution of the JC design problem.Based on the JC design,a constellation mapping method is presented.Compared with two conventional multi-access method in VLC systems,i.e.,time division multiplex access(TDMA)and non-orthogonal multiple access(NOMA),the proposed JC design obtain a better performance on symbol error rate.5.Autoencoder(AE)based non-linear transceiver design.Considering imperfect CSI channel,deep learning techonology is introduced and an autoencoder based transceiver design is proposed.To deal with the constraints of color and signal amplitude,a peak-value constraint layer and an integrated loss function are proposed in the AE structure.The new structure of AE network can be suitable for VLC systems with different numbers of colors.In perfect CSI channel,the proposed method can achieve a similar performance with the conventional CSK method.In imperfect CSI channel,the proposed transceiver design outperforms conventional CSK method.