Correlation-aware Resource Allocation in Multi-Cell Networks

TL;DR

This paper introduces a three-step, correlation-aware resource allocation strategy for multi-cell FDMA networks that significantly reduces distortion by considering source correlation and inter-cell interference.

Contribution

It presents a novel three-step cross-layer resource allocation algorithm that accounts for source correlation and inter-cell interference in multi-cell networks.

Findings

Achieves up to 37% higher distortion when correlation is used in resource allocation.

Demonstrates a 60% reduction in 5th percentile distortion compared to independent methods.

Shows the effectiveness of the proposed method through simulations in Gaussian source networks.

Abstract

We propose a cross-layer strategy for resource allocation between spatially correlated sources in the uplink of multi-cell FDMA networks. Our objective is to find the optimum power and channel to sources, in order to minimize the maximum distortion achieved by any source in the network. Given that the network is multi-cell, the inter-cell interference must also be taken into consideration. This resource allocation problem is NP-hard and the optimal solution can only be found by exhaustive search over the entire solution space, which is not computationally feasible. We propose a three step method to be performed separately by the scheduler in each cell, which finds cross-layer resource allocation in simple steps. The three- step algorithm separates the problem into inter-cell resource management, grouping of sources for joint decoding, and intra- cell channel assignment. For each of the steps we propose allocation methods that satisfy different design constraints. In the simulations we compare methods for each step of the algorithm. We also demonstrate the overall gain of using correlation-aware resource allocation for a typical multi-cell network of Gaussian sources. We show that, while using correlation in compression and joint decoding can achieve 25% loss in distortion over independent decoding, this loss can be increased to 37% when correlation is also utilized in resource allocation method. This significant distortion loss motivates further work in correlation-aware resource allocation. Overall, we find that our method achieves a 60% decrease in 5 percentile distortion compared to independent methods.