⚡ electrical engineering

AFDM-FTN: A Spectrally Efficient Waveform for High-Mobility Communications

This paper proposes the AFDM-FTN waveform, a spectrally efficient solution for high-mobility communications, and develops a low-complexity joint channel estimation and data detection scheme (BEM-MLMP-JCED) that leverages basis expansion modeling and multi-layer message passing to achieve enhanced spectral efficiency and reduced computational complexity compared to conventional benchmarks.

Original authors: Xianle Dai, Qu Luo, Jianguo Li, Fabien Heliot, Shuangyang Li, Lixia Xiao, Pei Xiao

Published 2026-07-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Xianle Dai, Qu Luo, Jianguo Li, Fabien Heliot, Shuangyang Li, Lixia Xiao, Pei Xiao

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 send a secret message across a crowded, bouncy trampoline park. This park represents the wireless world for high-speed trains or satellites, where the ground is constantly shifting and shaking (high-mobility).

For a long time, the standard way to send messages here was like a group of dancers (OFDM) trying to stay perfectly in step. But when the trampoline shakes too hard, they trip over each other, and the message gets garbled. A newer, cooler dance called AFDM (Affine Frequency Division Multiplexing) was invented. It's like a dance that uses spinning moves to stay balanced even when the trampoline wobbles. It's great, but it still leaves a little bit of empty space between the dancers, wasting some of the trampoline's real estate.

The Big Idea: Squeezing the Dancers
The authors of this paper asked a bold question: What if we could squeeze the dancers closer together to fit more of them on the trampoline without them tripping?

They combined the spinning dance of AFDM with a trick called FTN (Faster-Than-Nyquist). Think of FTN as a "super-squeeze" technique. Usually, you have to leave a gap between messages to avoid them crashing into each other. FTN says, "Let's skip the gap!" It intentionally lets the messages overlap just a tiny bit, creating a controlled "traffic jam" that the receiver can untangle. This allows the system to pack way more data into the same amount of space, boosting the Spectral Efficiency (SE)—basically, getting more bits per second per Hertz.

The Problem: The Messy Overlap
Here's the catch: When you squeeze messages together (FTN) on a shaking trampoline (high-mobility channel), the mess gets complicated. The messages don't just bump into their neighbors; they get twisted by the shaking ground, too. It's like trying to solve a puzzle where the pieces are moving and sticking to each other at the same time.

The paper argues that you can't just use the old, simple tools (like the standard LMMSE detector) to fix this. Those tools are like trying to untangle a knot with a blunt spoon—they get confused by the complex overlap and the shaking, leading to a lot of errors. The authors explicitly show that these older methods suffer "pronounced performance degradation" when the squeezing gets tight.

The Solution: A Three-Layer Detective Team
To solve this, the team built a new receiver called AFDM-FTN with a special brain named BEM-MLMP-JCED. Let's break down how this detective team works using a three-layer analogy:

The Time-Domain Layer (The Ground): This layer looks at the shaking ground. The team uses a "Basis Expansion Model" (BEM), which is like having a map of the trampoline's bounces. Instead of guessing every single bump, they use a few key patterns to describe the whole shake, making the job much faster and easier.
The FTN Layer (The Overlap): This layer handles the "super-squeezed" messages. Since the messages are packed tight, this layer knows exactly how they overlap and helps separate them out.
The Transform Layer (The Dance Floor): This is where the spinning dance happens. It takes the messy signals and rotates them into a clearer view.

The magic happens because these three layers talk to each other in a loop. The detective team passes "belief messages" back and forth. First, they guess the ground's shake. Then, they use that guess to untangle the squeezed messages. Then, they use the now-clearer messages to get a better guess of the ground's shake. They keep doing this, refining their answer over and over, until they get it right.

What the Simulations Showed
The authors didn't just dream this up; they ran thousands of computer simulations to test it. Here is what they found:

Speed vs. Accuracy: When they squeezed the messages (using a compression factor of 0.8, meaning they packed them tighter), the new system kept the error rate almost the same as the unsqueezed version. In fact, at a specific error rate of 1.1 × 10⁻³, the new system was about 3.5 dB better than the old "spoon" method (LMMSE).
The "Perfect" Benchmark: Even though they didn't have a perfect crystal ball (perfect channel information), their new system got very close to the performance of a system that did know everything about the shaking ground.
Real-World Gains: When they counted the actual data throughput (taking into account the "guard" spaces needed to keep things safe), the new system delivered an extra 0.20 bit/s/Hz for QPSK and 0.39 bit/s/Hz for 16-QAM compared to the standard method. That's like getting an extra lane on a highway without building new roads.
Complexity: The best part? The new system is actually less computationally heavy than the old methods. It's a faster, smarter detective that doesn't need a supercomputer to solve the puzzle.

The Bottom Line
The paper suggests that by combining the spinning dance of AFDM with the tight packing of FTN, and then using a smart, three-layer detective team to untangle the mess, we can send much more data to high-speed vehicles and satellites without the signal breaking. It's a way to get more speed out of the same airwaves, proving that sometimes, a little bit of controlled chaos (the overlap) is exactly what you need to move faster.

Technical Summary: AFDM-FTN: A Spectrally Efficient Waveform for High-Mobility Communications

Problem Statement
Next-generation wireless systems (6G and beyond) require enhanced spectral efficiency (SE) and ultra-high reliability, particularly in high-mobility scenarios such as high-speed railways and low-Earth-orbit (LEO) satellites. In these environments, channels are doubly selective (time-varying and frequency-selective), causing severe inter-carrier interference (ICI) in conventional Orthogonal Frequency-Division Multiplexing (OFDM). While Affine Frequency Division Multiplexing (AFDM) has emerged as a robust alternative due to its Doppler resilience, it traditionally adheres to the Nyquist criterion, limiting its SE. Conversely, Faster-Than-Nyquist (FTN) signaling improves SE by intentionally introducing controlled inter-symbol interference (ISI) but faces significant detection challenges when coupled with the time-varying nature of high-mobility channels. The core problem addressed is how to integrate FTN signaling into the AFDM framework (creating AFDM-FTN) to achieve high SE without incurring prohibitive computational complexity or performance degradation due to the complex coupling of channel time-variation, FTN-induced ISI, and data detection.

Methodology
The paper proposes a novel AFDM-FTN system and a corresponding low-complexity receiver architecture. The methodology proceeds through the following steps:

System Modeling: The authors derive the input-output relationship for AFDM-FTN. They demonstrate that FTN signaling introduces additional interference terms into the effective channel matrix, creating a coupled structure where multipath components overlap with FTN-induced ISI. The effective channel matrix is shown to be dense, complicating direct detection.
Channel Estimation (BEM): To address the challenge of estimating time-varying channels in the presence of FTN-ISI, the authors develop a Basis Expansion Model (BEM)-assisted channel estimator. This approach reformulates the nonlinear coupling between path coefficients and Doppler shifts into a compact linear model. By exploiting the intrinsic characteristics of the AFDM channel and FTN matrices, the method reduces the estimation complexity from $O(LN)$ to $O(L(Q+1))$ , where $L$ is the number of paths and $Q$ is the BEM order, thereby avoiding excessive pilot overhead and ill-conditioned issues.
Data Detection (MLMP): Recognizing that the dense effective channel matrix hinders standard message passing, the authors propose a Multi-Layer Message Passing (MLMP) algorithm. This detector factorizes the system into three layers:
- Time-Domain (TD) Channel Layer: Exploits the sparsity of the physical channel delay taps.
- FTN Layer: Exploits the sparsity of the known FTN coefficient matrix.
- Transform Layer: Utilizes the unitary nature of the Discrete Affine Fourier Transform (DAFT).
  Belief messages are iteratively propagated across these layers to perform low-complexity detection.
Joint Iterative Scheme (BEM-MLMP-JCED): The paper integrates the BEM estimator and the MLMP detector into a joint iterative scheme. In this framework, soft data estimates from the MLMP detector are fed back to refine the channel estimation, which in turn improves subsequent detection. This process iterates until convergence.

Key Contributions

AFDM-FTN System Proposal: A novel waveform that compresses the DAFT-domain signal duration and incorporates FTN signaling via practical pulse-shaping, specifically designed for high-mobility scenarios. The paper provides a rigorous derivation of the input-output relationship and analyzes the resulting interference patterns.
BEM-Assisted Estimation Framework: A low-complexity channel estimation method that linearizes the nonlinear coupling of channel parameters. It significantly reduces pilot overhead and computational burden compared to conventional methods.
MLMP Detector: An advanced detection algorithm that propagates messages across the TD channel, FTN, and transform layers. It explicitly exploits the sparsity of both the physical channel and the FTN coefficient matrix to enable efficient detection in a dense interference environment.
Joint Iterative Scheme (BEM-MLMP-JCED): A unified receiver that jointly performs channel estimation and data detection, achieving superior Bit Error Rate (BER) performance with reduced complexity.
Theoretical Analysis: The paper derives a lower bound on channel estimation error, analyzes the Mean Square Error (MSE) performance using state evolution to characterize convergence behavior, and provides a detailed computational complexity analysis.

Results
Simulation results validate the proposed system under various conditions:

Performance vs. Perfect CSI: The BEM-MLMP-JCED scheme converges within approximately 20 iterations, achieving BER performance close to that of a system with perfect Channel State Information (CSI).
Robustness to Compression: The system maintains comparable BER to conventional AFDM (Nyquist case) even with aggressive FTN compression factors (e.g., $\alpha = 0.8$ ), while providing significant SE gains.
Comparison with Benchmarks: The proposed AFDM-FTN with BEM-MLMP-JCED outperforms OFDM-FTN and OTFS-FTN schemes. Specifically, it achieves approximately a 3.5 dB gain over LMMSE detectors at a BER of $1.1 \times 10^{-3}$ for $\alpha = 0.8$ .
Spectral Efficiency: In terms of practical effective SE (accounting for pilot overhead, guard intervals, and BER), the proposed system with $\alpha = 0.8$ achieves an additional 0.20 bit/s/Hz (QPSK) and 0.39 bit/s/Hz (16-QAM) compared to conventional AFDM at high SNRs.
Complexity: The proposed algorithm exhibits linear complexity scaling with the number of subcarriers ( $O(N)$ ), offering a substantial improvement over the cubic complexity ( $O(N^3)$ ) of conventional LMMSE approaches.

Significance
The paper claims that the proposed AFDM-FTN system is particularly attractive for high-mobility channels because it successfully combines the Doppler resilience of AFDM with the spectral efficiency gains of FTN signaling. The significance lies in the development of a structured, low-complexity receiver (BEM-MLMP-JCED) that can handle the joint challenges of time-varying channels and controlled ISI. By demonstrating that high SE can be achieved without sacrificing BER performance or incurring prohibitive computational costs, the work provides a viable pathway for next-generation high-mobility communication systems. The theoretical analysis of the channel estimation lower bound and state evolution further provides insights into the fundamental limits and convergence behavior of the proposed architecture.

More like this