Abstract
We present a new spatiotemporal offloading (STO) framework that can deal with the time-varying traffic demand in an overloaded location using proactive caching during off-peak times. Focusing on a conventional cellular network, the system-wide mobility can be captured by the inter-cell transitions along individual users' mobility pattern. It is opposed to the existing geographical caching for content placement during off-peak times, which treats a system built around multiple base stations as a single service area, thus limiting its performance to low hit probability. We formulate a system-level STO problem to determine a content placement strategy that minimizes the access link load by delivering the requested data through device-to-device (D2D) links, rather than through a cellular link, particularly targeting potentially overloaded cells. Because its complexity is seriously attributed to the individual users' cell-by-cell mobility and content preferences, an aggregated flow model has been proposed to reduce the search space by invoking the law of segment conservation. Simulation results clearly demonstrate the effect of the spatiotemporal offloading that can be realized through inter-cell mobility-aware proactive caching. An extremely important implication of the current framework is that a varying traffic demand can be spatiotemporally distributed in a cellular system, overcoming the traditional limitation that arises from conservative or inefficient resource allocation due to a reliance on over-dimensioning.