⚡ electrical engineering

Failure-Aware Access Point Selection for Resilient Cell-Free Massive MIMO Networks

This paper proposes a Failure-Aware Access Point Selection (FAAS) method for cell-free massive MIMO networks that jointly optimizes channel strength and AP failure probabilities to significantly reduce outage rates and improve worst-case user performance under hardware failure conditions.

Original authors: Mostafa Rahmani Ghourtani, Junbo Zhao, Yi Chu, Hamed Ahmadi, David Grace, Alister G. Burr

Published 2026-02-19

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Mostafa Rahmani Ghourtani, Junbo Zhao, Yi Chu, Hamed Ahmadi, David Grace, Alister G. Burr

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The "Smart Waiter" Strategy for a Failing Restaurant

Imagine a massive, high-tech restaurant called Cell-Free Massive MIMO. Instead of having a few big dining rooms (traditional cell towers), this restaurant has hundreds of tiny, scattered waiters (Access Points or APs) spread all over the floor. Every waiter can serve every customer (User) at the same time.

The goal is to make sure every customer gets their food (data) quickly and reliably, even if the restaurant is huge and crowded.

The Problem: The "Sick Waiter" Scenario

In a perfect world, all the waiters are healthy and ready to work. But in reality, waiters get sick, run out of coffee, or trip over cables (hardware failures).

The Old Way (Failure-Agnostic): The manager assigns customers to waiters based only on who is standing closest. If a customer is assigned to a waiter who suddenly gets sick, that customer goes hungry. The manager didn't check the waiter's health first.
The Result: When many waiters get sick at once, some customers are left with no one to serve them, and the whole restaurant slows down.

The Solution: FAAS (Failure-Aware Access Point Selection)

The authors of this paper propose a new management strategy called FAAS. Think of it as a "Smart Waiter" algorithm that doesn't just look at who is closest, but also checks who is most likely to stay healthy.

Here is how it works, using simple analogies:

1. The "Health Score" (Failure Probability)
Every waiter has a "Health Score." Some are prone to tripping (high failure probability), while others are rock-solid (low failure probability).

The Stress Parameter ( $\alpha$ ): Imagine a dial the manager can turn.
- Dial at 0: It's a calm day; no one is sick.
- Dial at 1: It's a chaotic day; the building is shaking, and many waiters are likely to fall.
- The FAAS system adjusts its strategy based on how high this dial is turned.

2. The "Smart Assignment" Rule
When a customer needs food, the manager doesn't just pick the closest waiter. They pick a team of waiters who are:

Close enough to serve well (good signal).
Healthy enough to stay on their feet (low failure risk).

If the "Stress Dial" is high, the manager might skip a waiter who is very close but looks shaky, and instead pick a slightly further waiter who is very sturdy.

3. The "Safety Net" (Redundancy)
The system has a strict rule: Never assign a customer to just one waiter.

If a customer is assigned to a team of 3 waiters, and one gets sick, the other two keep serving.
The old method might have assigned a customer to just one waiter. If that one waiter fell, the customer was stranded.
FAAS ensures there is always a backup plan.

The Results: Why It Matters

The researchers ran simulations (virtual tests) to see how this new strategy performs compared to the old "closest waiter" method.

The "Worst-Case" Customer: In the old system, when things went wrong, the customers at the edge of the restaurant (or those assigned to sick waiters) got terrible service. With FAAS, even the worst-off customers got a decent meal.
The "Outage" Rate: This is the percentage of customers who got zero food because all their assigned waiters failed.
- The Old Way: When stress was high, many customers were left hungry (high outage).
- The FAAS Way: It reduced the number of hungry customers by over 85%. It's like having a safety net that catches almost everyone who would have fallen.

The "Crowded Room" vs. "Strong Waiters"

The paper also found something interesting about the layout of the restaurant:

Many Small Waiters (Single-Antenna APs): Having 400 small, simple waiters spread out was more resilient than having 100 super-strong waiters. Why? Because if one small waiter falls, there are 399 others nearby to help. It's like having a crowd of people holding an umbrella; if one lets go, the others hold it up.
Few Big Waiters (Multi-Antenna APs): If you rely on fewer, stronger waiters, and one of them gets sick, the gap is harder to fill.

The Bottom Line

This paper introduces a "smart manager" for future wireless networks. Instead of blindly assigning tasks based on distance, this manager looks at the reliability of the equipment. By proactively choosing the healthiest, most reliable "waiters" and ensuring there are always backups, the network stays online and fast, even when things go wrong.

It turns a fragile system into a resilient one, ensuring that even in a storm of hardware failures, your phone stays connected.

1. Problem Statement

Cell-Free Massive MIMO (CF-mMIMO) is a leading architecture for 6G, utilizing a large number of distributed Access Points (APs) to serve users without cell boundaries. While theoretically robust, existing literature often overlooks hardware resilience in the face of inevitable AP failures (caused by power outages, component degradation, or synchronization issues).

The Challenge: In traditional CF-mMIMO designs, user-to-AP associations are often static or based solely on channel strength (failure-agnostic). If an assigned AP fails, the user may experience service outage or severe spectral efficiency degradation.
The Gap: There is a lack of systematic frameworks that integrate probabilistic hardware failure models into the AP selection process to proactively maintain network reliability.

2. Methodology: Failure-Aware Access Point Selection (FAAS)

The authors propose FAAS, a dynamic selection strategy that jointly optimizes channel quality and AP reliability.

A. System Model

Network: A CF-mMIMO system with $M$ distributed APs and $K$ users.
Failure Model: Each AP $m$ has a baseline failure probability $p_{m,0}$ . A tunable stress parameter $\alpha \in [0, 1]$ scales this probability ( $p_m = \alpha p_{m,0}$ ) to simulate varying network stress levels. Failures are modeled as independent Bernoulli events.
Channel Estimation: Uses Minimum Mean Square Error (MMSE) estimation based on orthogonal pilot sequences.

B. The FAAS Algorithm

Unlike conventional schemes that select APs based purely on large-scale fading coefficients ( $\beta$ ), FAAS introduces a reliability-weighted selection metric:

Reliability Weighting: The selection metric for an AP $m$ serving user $k$ is defined as $\tilde{\beta}_{mk}(1 - p_m)$ , where $(1 - p_m)$ represents the survival probability.
Selection Rule: The Central Processing Unit (CPU) selects a subset of APs ( $\mathcal{M}_k$ ) such that the cumulative reliability-weighted gain meets a threshold $\varepsilon$ .
Redundancy Constraint: To prevent single points of failure, the algorithm enforces a minimum cooperative cluster size (e.g., at least 2 APs per user). This ensures that if one AP fails, the user remains connected to others.
Adaptability: As $\alpha$ (stress level) increases, the algorithm automatically shifts preference toward APs with higher survival probabilities, even if their nominal channel strength is slightly lower.

C. Resilience Metrics

The paper evaluates performance using two key metrics:

Minimum-User Spectral Efficiency (SE): The worst-case throughput among all users ( $\min_k E[SE_k]$ ). This ensures fairness and prevents marginalizing users with poor links.
User Outage Probability: The fraction of users who lose all active APs in their cluster due to failures.

3. Key Contributions

Systematic Integration of Failure Resilience: This is the first work to systematically integrate probabilistic hardware failure models into CF-mMIMO AP selection, bridging the gap between qualitative resilience discussions and quantitative system analysis.
FAAS Framework: Proposes a tractable, tunable algorithm that uses failure probabilities (from monitoring or historical data) to proactively optimize user-AP associations.
Theoretical and Practical Validation: Provides a framework for analyzing the impact of independent AP failures on spectral efficiency and outage probability, moving beyond closed-form analytical complexity to a simulation-based evaluation of realistic stress scenarios.
Topology Agnosticism: The proposed selection logic relies only on local channel and reliability data, making it applicable to star, hierarchical, or O-RAN disaggregated architectures.

4. Numerical Results

The authors simulated a 2x2 km network with 100 users under two configurations:

High Density: 400 single-antenna APs ( $N=1$ ).
Low Density: 100 four-antenna APs ( $N=4$ ).

Key Findings:

Performance under Stress ( $\alpha=1$ ):
- Outage Reduction: FAAS reduces user outage probability by over 85% compared to failure-agnostic clustering at high failure intensities.
- Spectral Efficiency: FAAS maintains significantly higher minimum-user spectral efficiency. While failure-agnostic schemes suffer sharp drops in the lower tail of the CDF (worst-case users), FAAS preserves performance by avoiding unreliable APs.
Comparison of Configurations:
- The single-antenna AP configuration (400 APs) demonstrated superior resilience compared to the multi-antenna configuration (100 APs).
- Reason: The dense deployment of single-antenna APs offers greater macro-diversity. Users with weak links benefit more from the broader spatial distribution of many APs than from the local interference suppression of fewer, more capable APs.
Baseline Comparison:
- "All Available APs" (idealized, no failures) performs best but is unscalable.
- "Failure-Agnostic" performs well at $\alpha=0$ but degrades rapidly as failures occur.
- FAAS strikes the optimal balance, adapting to stress conditions to maintain robustness.

5. Significance

This paper establishes a critical foundation for resilient 6G network design.

Practicality: It moves beyond theoretical assumptions of perfect hardware, addressing real-world issues like power outages and component faults.
Design Paradigm: It demonstrates that "resilience-by-design" is achievable in CF-mMIMO by simply weighting AP selection with failure probabilities.
Future Impact: The results suggest that future CF-mMIMO deployments should prioritize dense, distributed AP architectures (even with single antennas) to maximize macro-diversity and fault tolerance. The framework also opens avenues for optimizing networks based on quantile-based guarantees (e.g., 99% reliability) rather than just average performance.

In summary, FAAS provides a practical, efficient solution to ensure that next-generation wireless networks can deliver ubiquitous, reliable connectivity even in the presence of inevitable hardware failures.