摘要： Due to the limited battery capacity in mobile devices, simultaneous wireless information and power transfer (SWIPT) has been proposed as a promising solution to improving the energy efficiency (EE) in Internet of things (IoT) networks, i.e., device-to-device (D2D) networks, by allowing mobile devices to harvest energy from ambient radio frequency (RF) signals. However, the non-linear behavior of RF energy harvesters has largely been ignored in existing works on SWIPT. In this paper, we propose to maximize the sum EE of all D2D links in a D2D underlaid cellular network by optimizing the resource and power allocation based on a non-linear energy harvesting (EH) model. Toward this end, we first propose a pre-matching algorithm to divide the D2D links into a SWIPT enabled group and a non-EH group that cannot meet the EH circuit sensitivity. We then develop a two-layer iterative algorithm to jointly optimize the D2D transmission power and the power splitting ratio to maximize the EE for each SWIPT enabled D2D link. On this basis, we build the preference lists for both SWIPT enabled D2D links and cellular user equipment (CUEs), and propose a one-to-one constraint stable matching algorithm to maximize the sum EE of all SWIPT enabled D2D links by optimizing the spectrum resource sharing between D2D links and CUEs. The sum EE of non-EH D2D links is maximized through an iterative power control algorithm and a one-to-one stable matching algorithm. Simulation results show that our proposed algorithms achieve a much higher sum EE than the existing matching based energy-efficient resource allocation scheme for SWIPT enabled D2D networks.