Research On Time-domain Interference Coordination And Resource Allocation In Heterogeneous Networks

With the rapid development of mobile communication and the surge of Internet application, there are plentiful data service categories and boosting service traffic, and users have a higher demand for Quality of Service (QoS). As a result, heterogeneous networks are proposed, with low power nodes introduced in the traditional macro deployment, to enhance users’signal to interference and noise ratio and throughput by shortening the distance between base stations and users. In heterogeneous networks, inter-cell interference is the major interference source and different type of low power nodes can cause different interference problem. In Macro-Pico scenario with the spectrum shared, Pico is weak in coverage and offloading due to its low power. So3GPP (3rd Generation Partnership Project) has proposed Cell Range Expansion (CRE) to enhance Pico coverage and achieve load balancing. Morever, time-domain inter-cell interference coordination (ICIC) based on Almost Blank Subframe (ABS) which includes zero power ABS and low power ABS is proposed to decrease Pico CRE users’interference. In addition, resouce management has a tight relationship with interference problem, which means efficient resource allocation, scheduling and power control schemes can reduce the interference, satisfying users’QoS demand and improving the network utility.In this dissertation, we focus on time-domain interference coordination and resource allocation in Macro-Pico scenario. The contributions are as follows.1) Considering Macro ABS power control and network resource allocation problem, we formulate it as an optimization problem to maximize network throughput utility under the constraints of physical resource and users’data rate demands. A joint Resource Allocation and ABS Power Control (RA-APC) algorithm is proposed, which results in a local optimal solution by iterating between resource allocation and ABS power control. The proof of convergence is also demonstrated. Simulation results show that the algorithm converges fast and can better utilize Macro time-domain resource and achieve better network utility compared to zero power ABS.2) In addition to ABS power control, each Pico should select its own CRE bias according to different coverage owing to its position and different user distribution, especially when users move. Considering Pico CRE bias and Macro ABS power control jointly, as well as QoS demands of different services, we propose a Q learning based time-domain ICIC and resource allocation scheme. In the Q learning scheme, base stations, acting as agents, interact with the environment and learn to select related network metrics like CRE bias, ABS low power to approximate the optimal strategy by updating the Q tables. Simulation results show that selecting network metrics based on Q learning can improve system utility.