Active IRS-Assisted Wireless Powered MEC Systems with Non-linear Energy Harvesting Model.

Intelligent reflecting surface (IRS) is a promising solution to assist the wireless power transfer (WPT) and task offloading in wireless powered mobile edge computing (WP-MEC) systems. However, to alleviate the double-fading attenuation of IRS-involved links, passive IRS is conventionally equipped with a large number of reflecting elements (REs), resulting in large surface size. To solve this issue, we investigate an active IRS-assisted WP-MEC system, in which active IRS is deployed to amplify the incident signals instead of just adjusting phase shifts as passive IRS. The beamforming at multi-antenna hybrid access point (HAP) is utilized for further performance improvement. To capture the sensitive properties of energy harvesting (EH) circuits, a practically non-linear EH model is adopted. The tasks are assumed to be bit-wise independent, and thus the partial offloading policy is adopted to the wireless devices (WDs). Accordingly, we propose an efficient compute-while-offload mode, i.e., the offloaded tasks are executed as soon as they are received by the HAP, thereby making full use of the finite edge computing capability. We aim to maximize the sum computational rate (SCR) of all the WDs by jointly optimizing time allocation, beamforming designs of active IRS and HAP, as well as offloading strategies of WDs. An algorithm based on alternating optimization (AO) is proposed to tackle the non-convexity of this problem. Numerical results demonstrate the superiority of deploying active IRS over passive IRS.