Online Resource Allocation for Energy Harvesting Downlink Multiuser Systems: Precoding With Modulation, Coding Rate, and Subchannel Selection

• Published in: IEEE transactions on wireless communications Vol. 14; no. 10; pp. 5780 - 5794
• Main Authors: Weiliang Zeng, Zheng, Yahong Rosa, Schober, Robert
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.10.2015; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Allocations; Coding; Dynamic programming; Energy consumption; Energy harvesting; finite alphabet inputs; Internet resources; Joints; Mathematical models; MIMO; Modulation; multiuser systems; online resource allocation; Optimization; Resource management; statistical channel state information; stochastic dynamic programming; Transmitters; Wireless communication; statistical channel state information; finite-alphabet inputs; online resource allocation; multiuser systems; Energy harvesting; stochastic dynamic programming
• ISSN: 1536-1276; 1558-2248
• DOI: 10.1109/TWC.2015.2442987

This paper proposes an online resource allocation algorithm for weighted sum rate maximization in energy harvesting downlink multiuser multiple-input-multiple-output (MIMO) systems, where the base station transmitter is powered by both a regular energy source and an energy buffer that is connected to an energy harvester. Taking into account the discrete nature of the modulation and coding rates (MCRs) used in practice, we formulate a stochastic dynamic programming (SDP) problem to jointly design the MIMO precoders, select the MCRs, assign the subchannels, and optimize the energy consumption over multiple time slots with causal and statistical energy arrival information and statistical channel state information. Solving this high-dimensional SDP entails several difficulties: the SDP has a nonconcave objective function, the optimization variables are of mixed binary and continuous types, and the number of optimization variables is on the order of thousands. We propose a new method to solve this NP-hard SDP by decomposing the high-dimensional SDP into an equivalent three-layer optimization problem and develop efficient algorithms to solve each layer separately. The decomposition reduces the computational burden and breaks the curse of dimensionality successfully. We analyze the complexity of the proposed algorithm and demonstrate the performance gains based on numerical examples.