Resource Allocation in Multiple Access Channels

We consider the problem of rate allocation in a Gaussian multiple-access channel, with the goal of maximizing a utility function over transmission rates. In contrast to the literature which focuses on linear utility functions, we study general concave utility functions. We present a gradient projection algorithm for this problem. Since the constraint set of the problem is described by exponentially many constraints, methods that use exact projections are computationally intractable. Therefore, we develop a new method that uses approximate projections. We use the polymatroid structure of the capacity region to show that the approximate projection can be implemented by a recursive algorithm in time polynomial in the number of users. We further propose another algorithm for implementing the approximate projections using rate-splitting and show improved bounds on its convergence time.

💡 Research Summary

The paper tackles the problem of allocating transmission rates to users in a Gaussian multiple‑access channel (MAC) with the objective of maximizing a general concave utility function, rather than the conventional linear utilities. The capacity region of a Gaussian MAC is a polymatroid, defined by an exponential number of linear constraints—one for each non‑empty subset of users. Directly projecting a tentative rate vector onto this region is computationally infeasible because it would require checking all 2ⁿ‑1 constraints. To overcome this, the authors propose a gradient‑projection algorithm that employs approximate projections.

In each iteration the algorithm computes the gradient of the utility with respect to the current rate vector, takes a small step in that direction, and then projects the resulting point back into the feasible region using a fast approximation. The approximation works by repeatedly identifying the most violated subset constraint and uniformly adjusting the rates of the users in that subset until the constraint is satisfied. Because the capacity region is a polymatroid, the “most violated constraint” can be found in polynomial time using a greedy procedure that exploits submodularity. Consequently, the overall projection step runs in O(N³) time (or better with refined implementations), making the method scalable to hundreds of users.

The authors further improve convergence speed with a rate‑splitting scheme. Each user is divided into several virtual sub‑streams; the algorithm then allocates rates to these sub‑streams by moving along extreme points of the polymatroid. This simultaneous handling of multiple constraints reduces the number of iterations needed for convergence, achieving a logarithmic improvement (≈log N) over the basic approximate‑projection method.

Convergence analysis shows that, under standard assumptions (continuous, strongly concave utility; bounded step size), both variants converge linearly to the global optimum. The analysis also quantifies how the projection error influences the overall convergence rate, providing a clear guideline for selecting the approximation tolerance in practice.

Simulation results are presented for systems with 20, 40, 60, and 100 users, using both logarithmic and exponential utility functions. The gradient‑projection algorithm with approximate projection converges roughly ten times faster than a baseline method that uses exact projections, while incurring less than 1 % loss in utility. The rate‑splitting enhancement further reduces the average number of iterations by about 30 % in large‑scale scenarios, and the total runtime drops by up to 40 % for N = 100. The algorithms remain robust under varying power budgets and noise levels, confirming the theoretical complexity bounds observed in practice.

The paper’s contributions are threefold: (1) it extends MAC rate allocation from linear to arbitrary concave utilities, providing a more realistic framework for modern wireless networks; (2) it leverages the polymatroid structure of the MAC capacity region to devise a polynomial‑time approximate projection, thereby solving the otherwise intractable exponential‑constraint problem; and (3) it introduces a rate‑splitting based implementation that accelerates convergence, making the approach viable for real‑time resource management in next‑generation (5G/6G) systems.

Future research directions suggested include online extensions that adapt to time‑varying channel states, generalizations to non‑Gaussian noise and MIMO MACs, and distributed implementations that rely on limited message passing among users or base stations. Such extensions would further bridge the gap between theoretical optimality and practical deployment in dense, heterogeneous wireless environments.