SWIPT-based Real-Time Mobile Computing Systems: A Stochastic Geometry Perspective

Abstract

Driven by the Internet of Things vision, recent years have seen the rise of new horizons for the wireless ecosystem in which a very large number of mobile low power devices interact to run sophisticated applications. The main hindrance to the massive deployment of low power nodes is most probably the prohibitive maintenance cost of battery replacement and the ecotoxicity of the battery production/end-of-life. An emerging research direction to avoid battery replacement is the combination of radio frequency energy harvesting and mobile computing (MC). In this paper, we propose the use of simultaneous information and power transfer (SWIPT) to control the distributed computation process while delivering power to perform the computation tasks requested. A real-time MC system is considered, meaning that the trade-off between the information rate and the energy harvested must be carefully chosen to guarantee that the CPU may perform tasks of given complexity before receiving a new control signal. In order to provide a system-level perspective on the performance of SWIPT-MC networks, we propose a mathematical framework based on stochastic geometry to characterise the rateenergy trade-off of the system. The resulting achievable performance region is then put in relation with the CPU energy consumption to investigate the operating conditions of real-time computing systems.Finally, numerical results illustrate the joint effect of the network densification and the propagation environment on the optimisation of the CPU usage. Index TermsSimultaneous wireless information and power transfer (SWIPT), RF energy harvesting, mobile computing (MC), trade-off, stochastic geometry, network analysis. A. I. Akin, N. Janatian, I. Stupia and L. Vandendorpe are with the about the possible limitations in energy storage and computing capacity for the mobile low power devices (LPDs) that will populate the future wireless ecosystem. Currently, researchers and engineers are striving to lay the foundations of a technological breakthrough towards a new generation of LPDs with enhanced real-time processing capability that would support the IoT vision. In this sense, the appealing solutions proposed so far delineate a new framework for the design and the optimisation of wireless networks in which computation and traditional communication aspects are merged into a unified perspective [1]. The consequence is a new network formalism in which the final objective of the design process is not the maximisation of the transferable information per se, but rather the maximisation of the information instrumental in guaranteeing an optimal use of the limited computation resources available at the LPDs [2]. Following this line of thought, it appears that the implementation of such mobile computing (MC) systems is confronted with a double bottleneck: i) the restrictions on the amount of energy available to perform sophisticated computation tasks, and ii) the limited capacity of the radio links. A potential solution to overcome this double obstacle consists in a ...