Fast Power Allocation Algorithm Based On Optimal Energy Efficient Multi-Channel Wireless Communication Transmissions

We consider power allocation in energy efficient multi-channel wireless OFDM(Orthogonal Frequency Division Multiplexing) or MIMO (Multiple-InputMultiple-Output) transmissions. An energy-per-goodbit (EPG) metric is used to gaugethe average energy consumed for transmitting each bit. Along with the rapid growth ofradio technologies and services, there is a growing concern on how to reduce energycost and increase energy efficiency in wireless communications. This introduces acommon objective for system optimization, in some scenarios it may be moreappropriate to maximize the information bits sent per unit energy, or reciprocallyminimize an energy-per-goodbit (EPG) metric. The main work of this paper is to obtaina power allocation vector minimizing EPG.Existing works minimize EPG by searching for a set of dual variables which areused to compute the optimal power allocation. In multi-channel wirelesscommunications, Water-Filling is a fundamental power allocation mechanism forcapacity maximization under a given total transmits power. According to the parallelproperty of WF, we propose simple and fast algorithms to compute optimal WFsolutions. Both scenarios, with and without an upper-bound on the allocated powers inthe sub-channels, are considered. Unlike the conventional WF approach which needs tosearch for a Lagrange multiplier (i.e., the water level), our algorithms require much lesscomputations than the existing ones by removing the water-level searching process.They can converge to the optimal solutions multiple times faster in a few linearcalculations, such that the system can better match the dynamic nature of the wirelesschannels for a better performance. Our fast WF algorithms are significantly importantfor solving energy efficient optimization.The Existing searching process is complex and requires a long running time. In thispaper, we propose a new algorithm to compute the optimal power allocation withoutsearching any dual variables. Our algorithm is based on an iterative calculation of thetotal transmission power and it converges when the minimum EPG is reached. It takesWF as the basic building block. In the meantime, we study the property of the EPG function and verify its convexity, which is so imperative for our proposed algorithm.Then, the convergence and optimality of our iteration algorithm is proved. The onlybuilding block of energy efficient algorithm is our modified fast and effective WFalgorithm requiring much less running time. Numerical results show that our powerallocation algorithm for EPG minimization, together with the embedded fast WFalgorithm, can run multiple times faster than the existing ones.