Power Allocation Policy Over Multiple Wireless Channels

With the advantage of convenience and mobility, wirelesscommunication has penetrated into our working, academic research anddaily lives. However, multiple impairments suffered by wirelesscommunication systems, for example, path loss, mutual interferenceamong channels and limited power supply of devices, hindered thedevelopment of new applications of wireless network, making it more andmore important to implement adaptive power allocation technology. In thisthesis, we consider the problem of optimal power allocation over multiplewireless channels. We first proposed an abstract model with a number ofindependent and identically distributed Gilbert Elliot channels, and thenwe designed a general power allocation strategy which distributes thelimited transmission power evenly to good quality channels only.Considering the fact that channel state information is not known whenpower allocation decision is made, we introduced the concept of belief thatrepresents the probability that a channel will be in good state, and usedPartially Observable Markov Decision Process to theoretical formulate theproblem. In order to maximize the long-term expected discountedtransmitted data, we developed the value function and tried to find anoptimal policy which is defined as an optimal mapping from believes topower allocation actions. We theoretically derived some useful features ofthe value function and decision regions, based on which the overallstructure of the optimal policy is obtained. In the case of a three-channelsystem, we prove that the decision regions are single-connected regionsextended from the8vertices of belief space. We then carried outsimulations based on linear programming and AMPL. The simulation results verified our derived structure and showed how to calculate theoptimal policy once the channel parameters are given.