⚡ electrical engineering

Semi-Blind Receivers for RIS-Aided Fluid Antenna Systems

This paper proposes a semi-blind estimation framework using tensor models (PARAFAC and Nested PARAFAC2) to concurrently estimate channels and symbols in RIS-aided fluid antenna systems, effectively reducing training overhead while offering a trade-off between robustness and computational complexity.

Original authors: Josué V. de Araújo, Gilderlan T. de Araújo, André L. F. de Almeida

Published 2026-04-21

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Josué V. de Araújo, Gilderlan T. de Araújo, André L. F. de Almeida

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

Imagine you are trying to have a conversation with a friend in a crowded, noisy room. You can't shout, and you can't move around freely. This is a bit like how current wireless networks (like 5G) work: they send signals through fixed antennas that are stuck in one spot, trying to find a clear path through the "noise" of the city.

This paper introduces a new, super-smart way to talk to your friend by combining two futuristic technologies: Reconfigurable Intelligent Surfaces (RIS) and Fluid Antennas (FAs).

Here is the breakdown of the problem and the paper's solution, using simple analogies.

The Problem: The "Blind" Conversation

In a normal wireless system, the phone (the user) sends a message to the tower (the base station). To do this well, the tower needs to know exactly how the signal travels—where the walls are, where the interference is, and how to aim the signal.

Usually, the phone has to send a "test signal" (like shouting "Hello? Can you hear me?") so the tower can map the room. This takes up time and energy.

The new twist:

RIS (The Smart Mirror): Imagine a wall covered in thousands of tiny, programmable mirrors. These mirrors can instantly change the angle of the light (or radio signal) bouncing off them to steer it exactly where you want.
Fluid Antennas (The Moving Ear): Imagine the base station doesn't have fixed ears. Instead, it has a "fluid" antenna that can physically move its listening point to different spots within a small area, or switch between many different "ears" very quickly.

The Challenge:
Because both the mirrors (RIS) and the ears (Fluid Antenna) are constantly moving and changing, the path the signal takes is changing every millisecond. If the tower tries to map this path using old "test signals," it would have to shout "Hello" millions of times a second just to keep up. That's too slow and wastes too much energy. The tower is effectively "blind" because the map changes too fast to draw.

The Solution: The "Semi-Blind" Detective

The authors of this paper propose a clever trick. Instead of shouting "Hello" to get a map, the tower acts like a detective who can figure out the map just by listening to the actual conversation.

They call this a "Semi-Blind Receiver."

Blind: The tower doesn't need a pre-written script (pilot signals) to know what to listen for.
Semi-Blind: It uses a tiny bit of known information (just enough to start) and then uses math to figure out the rest.

The Two Strategies (Protocols)

The paper suggests two different ways to organize this "detective work," depending on how much time and computing power you have.

Protocol 1: The "Slow-Motion" Strategy (The PF Model)

How it works: Imagine the conversation happens in distinct "scenes." In each scene, the mirrors stay still for a moment, and the fluid antenna picks a specific spot to listen. Then, the mirrors change, and the antenna moves to a new spot for the next scene.
The Analogy: It's like taking a series of high-quality, slow-motion photos. Because the setup is very stable during each "photo," the detective can build a very detailed, 3D map of the room.
The Result: This method is extremely accurate. It recovers the message perfectly, almost as if the tower had perfect vision.
The Catch: It requires strict timing and more complex hardware to coordinate the "scenes." It's like needing a very expensive camera crew.

Protocol 2: The "Fast-Paced" Strategy (The NPF Model)

How it works: Imagine the conversation is a fast-paced movie. The mirrors and the antenna are changing positions constantly and rapidly, all at once.
The Analogy: Instead of slow-motion photos, the detective is watching a fast-forwarded video. The picture is a bit blurrier, but the detective uses a special mathematical trick (called a "Nested Tensor") to piece together the clues from the blur.
The Result: This method is faster and cheaper to run. It doesn't need as much coordination.
The Catch: It's slightly less accurate than Protocol 1, but it's much more flexible and easier to build into real devices.

The Secret Weapon: "Tensor Math"

You might wonder, "How does the detective solve the puzzle?"
The paper uses a branch of math called Tensor Decomposition.

The Analogy: Imagine you have a giant 3D block of Jell-O. Inside the block, there are invisible strings (the signals) running in three different directions: Up/Down, Left/Right, and Forward/Backward.
Normally, if you look at just one slice of the Jell-O, you can't see the whole picture. But Tensor math allows the detective to look at the entire block at once. It separates the "strings" from the "Jell-O" (the noise) by looking at how they interact in all three dimensions simultaneously.
This allows the tower to separate the user's voice from the background noise and figure out exactly where the signal bounced off the mirrors, even without a "Hello" test signal.

Why Does This Matter?

Speed: It saves time. The network doesn't need to stop and send test signals, so data flows faster.
Efficiency: It uses less battery power because it doesn't waste energy on constant "Hello" signals.
Future-Proofing: As we move toward 6G, we need to connect millions of devices. This system uses "spatial degrees of freedom" (moving the antenna and mirrors) to squeeze more data into the same amount of space, making our future networks much more powerful.

Summary

The paper says: "Don't try to map a moving room with a static map. Instead, use a smart detective (the receiver) and a special 3D math tool (Tensor Decomposition) to figure out the room's shape just by listening to the conversation. We offer two ways to do this: one that is super-accurate but complex (Protocol 1), and one that is fast and flexible (Protocol 2)."

This is a major step toward making our future wireless networks smarter, faster, and more energy-efficient.

1. Problem Statement

The paper addresses the critical challenge of channel acquisition in next-generation wireless networks that combine Reconfigurable Intelligent Surfaces (RIS) and Fluid Antennas (FAs).

The Challenge: In RIS-aided FA systems, the effective channel is a cascaded link (User $\to$ RIS $\to$ BS) that depends on two dynamic factors: the RIS phase-shifting configuration and the FA port selection (which antenna ports are active).
The Bottleneck: Conventional pilot-based channel estimation methods require excessive training overhead because the channel changes with every reconfiguration of the RIS and the FA ports. This makes the system inefficient for high-mobility or low-latency scenarios.
The Gap: Existing literature treats RIS channel estimation and FA channel estimation separately. There is no framework for joint semi-blind estimation (estimating channels and data symbols simultaneously without full pilot sequences) for a system combining both technologies.

2. Methodology

The authors propose a tensor-based semi-blind estimation framework that exploits the multi-dimensional structure of the received signal to decouple channels and symbols.

A. System Model

Architecture: A multi-user (MU) uplink where $K$ single-antenna users communicate with a Base Station (BS) equipped with a 1D Fluid Antenna array (selecting $M$ active ports out of $N$ available) via an RIS with $N_r$ elements.
Signal Structure: The received signal is modeled as a function of the user-to-RIS channel ( $G$ ), RIS-to-BS channel ( $H$ ), RIS phase shifts, and FA port selection matrices.

B. Two Transmission Protocols

The paper introduces two hierarchical protocols that result in different tensor models:

Protocol 1: Two-Time-Scale (TTS) Structure
- Mechanism: The FA port selection and RIS configuration remain constant over a block of time slots ( $P$ ) but change between spatial blocks ( $I$ ).
- Tensor Model: This structure yields a standard PARAFAC (PF) model. The factor matrices (spatial, temporal, coding) remain fixed across slices, allowing for a rigid, highly structured decomposition.
- Receiver: A PF Receiver based on Trilinear Alternating Least Squares (TALS).
Protocol 2: Single-Time-Scale (STS) Structure
- Mechanism: The FA port selection and RIS configuration change dynamically at every transmission block ( $I$ ), while the coding structure is applied within the block.
- Tensor Model: This leads to a Nested PARAFAC2 (NPF) model. Here, the spatial factor matrix varies across slices (blocks), and this varying factor itself follows a PARAFAC2 structure due to the cascaded nature of the RIS and FA.
- Receiver: An NPF Receiver based on TALS, designed to handle the slice-dependent variations.

C. Estimation Algorithm

For both protocols, the authors derive semi-blind receivers using Trilinear Alternating Least Squares (TALS):

Objective: Jointly estimate the user-to-RIS channel ( $G$ ), RIS-to-BS channel ( $H$ ), and transmitted symbols ( $X$ ) without prior knowledge of pilot sequences.
Process: The algorithm iteratively updates estimates of $G$ , $H$ , and $X$ by minimizing the Frobenius norm of the reconstruction error, leveraging the Kronecker product properties of the tensor unfoldings.

3. Key Contributions

First Joint Framework: This is the first work to propose a joint semi-blind estimation framework for a multi-user uplink combining RIS and Fluid Antennas.
Dual Protocols & Models: The introduction of two distinct signaling protocols (TTS and STS) mapped to PARAFAC and Nested PARAFAC2 tensor models, respectively.
Algorithm Development: Derivation of efficient TALS-based receivers for both models that exploit spatio-temporal diversity from RIS reconfiguration and FA port switching.
Theoretical Analysis:
- Identifiability Conditions: Established necessary conditions (e.g., $IMTP \geq N_r \max(K, N)$ for Protocol 1) to ensure unique channel recovery.
- Complexity Analysis: Provided computational complexity expressions, revealing a trade-off: Protocol 1 offers higher robustness but higher complexity, while Protocol 2 is lighter and more flexible.

4. Simulation Results

The authors validated their approach through extensive simulations:

Performance vs. Pilot-Assisted: Both semi-blind receivers achieve performance very close to the theoretical "Perfect CSI" and "Pilot-Assisted" bounds, significantly reducing training overhead.
Protocol Comparison:
- Protocol 1 (PF): Demonstrates superior Bit Error Rate (BER) and Normalized Mean Square Error (NMSE) performance, especially in mid-to-high SNR regimes. It converges faster to the ideal bound due to richer spatial diversity.
- Protocol 2 (NPF): Offers a lower-complexity alternative with slightly lower performance, suitable for scenarios with stricter hardware or synchronization constraints.
Resource Allocation: Results show that increasing spatial diversity (more blocks of RIS/FA reconfiguration) is more effective for performance than simply extending the temporal duration of symbols.
Scalability: The system scales effectively with the number of users ( $K$ ). The performance penalty of adding users can be compensated by increasing the number of spatial snapshots.
Spectral Efficiency (SE): Joint optimization of RIS phases and FA port selection yields the highest SE gains compared to using only RIS or only FAs.

5. Significance

Overcoming Training Overhead: The proposed semi-blind approach solves the "curse of dimensionality" in RIS-FA systems, where traditional pilot-based methods would require prohibitive training time.
Enabling 6G Technologies: It provides a practical signal processing foundation for integrating two key 6G enablers (RIS and FAs), demonstrating that their combined spatial degrees of freedom can be harvested efficiently.
Design Trade-offs: The paper offers a clear trade-off analysis for system designers: choose Protocol 1 for maximum accuracy and robustness (at the cost of complexity/synchronization), or Protocol 2 for a flexible, lower-complexity implementation.
Tensor Theory Application: It successfully applies advanced tensor decompositions (specifically Nested PARAFAC2) to a complex wireless communication problem, opening new avenues for signal processing in dynamic environments.