The recent flood of mobile data is motivating the deployment of dense wireless networks to boost the capacity of next-generation cellular networks. However, deploying more and more base stations (BSs) will impose heavy burdens on existing backhaul links. It could cause much congestion and long latency, which may end up degrading the user experience. The infrastructure update of backhaul links could also be labor-intensive and require considerable expenditures. Caching popular contents at BSs has recently been proposed as a powerful and cost-effective supplement to existing limited backhaul links for accommodating the tremendous amount of mobile data traffic. Cache-assisted communications enjoy important benefits, such as increasing successful transmission probability, reducing d...[ Read more ]

The recent flood of mobile data is motivating the deployment of dense wireless networks to boost the capacity of next-generation cellular networks. However, deploying more and more base stations (BSs) will impose heavy burdens on existing backhaul links. It could cause much congestion and long latency, which may end up degrading the user experience. The infrastructure update of backhaul links could also be labor-intensive and require considerable expenditures. Caching popular contents at BSs has recently been proposed as a powerful and cost-effective supplement to existing limited backhaul links for accommodating the tremendous amount of mobile data traffic. Cache-assisted communications enjoy important benefits, such as increasing successful transmission probability, reducing delivery time and backhaul traffic, and lowering the total network cost. To fully exploit the benefits of caching at the wireless edge, critical design issues need to be carefully addressed. The main theme of this thesis is to investigate key design problems in cache-assisted communications for backhaul-limited cellular networks.

To achieve a cost-efficient cache deployment, we first investigate the cache size allocation problem under a given budget by considering wireless channel statistics, backhaul conditions and file popularity distributions. The user success probability (USP) is proposed as the performance metric. For the single-cell scenario, a closed-form expression for the USP is derived. Analytical results will show that the minimum required cache size will increase exponentially when the USP threshold or the number of MUs gets larger, and will decrease exponentially when the backhaul capacity or the success probability of wireless transmission becomes higher. Moreover, the minimum required cache size is independent of the total number of files under a highly concentrated file popularity distribution, while it is in linear relation with the total number of files for a less concentrated file popularity distribution. We also study the cache size allocation in the multi-cell scenario and provide a bisection search algorithm to find the optimal solution.

When considering the prefetching phase and in order to properly utilize cache space, we study the caching placement for minimizing the average download delay. The optimal caching placement should be aware of the backhaul delays, wireless channel statistics and BS cooperation. The lower bound of the average download delay, which explicitly shows the impacts of key system parameters, is derived. The caching placement problem turns out to be a mixed-integer nonlinear programming (MINLP) problem and difficult to solve. To tackle this, we relax the problem into a DC programming problem and propose a low-complexity algorithm based on successive convex approximation. Simulation results shall show that the proposed low-complexity algorithm can achieve comparable performance to exhaustive search, and outperforms conventional strategies adopted in prior works.

In the delivery phase, we will propose to jointly design backhaul data assignment and unicast beamforming in caching networks. We aim at minimizing the network cost which is defined as a weighted sum of the backhaul cost and the transmit power cost, and consider partial BS coordination. The joint design is formulated as an MINLP problem and is highly complicated. In order to provide an efficient solution, group sparse optimization is applied by enforcing the sparsity structures in the backhaul data assignment and in the aggregate beamformer. Simulation results will demonstrate a significant reduction in the network cost by introducing caches, thereby confirming caching as a potential cost-effective technique in future wireless networks.

Finally, we will investigate a more challenging scenario which unifies active BS selection, backhaul data assignment and adaptive multicast beamforming to minimize the total network power consumption for cache-enabled wireless networks, while taking into account the quality-of-service (QoS) requirements and backhaul capacity limitations. We shall propose a three-stage layered group sparse beamforming (LGSBF) framework, which is a generalization of the power efficiency problems in prior works. Simulation results will validate the effectiveness of the proposed algorithm in reducing the network power consumption, and demonstrate that caching plays a more significant role in networks with higher user densities and less-power-efficient backhaul links.