Agile Affine Frequency Division Multiplexing

The advancement to 6G calls for waveforms that transcend static robustness to achieve intelligent adaptability. Affine Frequency Division Multiplexing (AFDM), despite its strength in doubly-dispersive channels, has been confined by chirp parameters optimized for worst-case scenarios. This paper shatters this limitation with Agile-AFDM, a novel framework that endows AFDM with dynamic, data-aware intelligence. By redefining chirp parameters as optimizable variables for each transmission block based on real-time channel and data information, Agile-AFDM transforms into an adaptive platform. It can actively reconfigure its waveform to minimize peak-to-average power ratio (PAPR) for power efficiency, suppress inter-carrier interference (ICI) for communication reliability, or reduce Cramer-Rao bound (CRLB) for sensing accuracy. This paradigm shift from a static, one-size-fits-all waveform to a context-aware signal designer is made practical by efficient, tailored optimization algorithms. Comprehensive simulations demonstrate that this capability delivers significant performance gains across all metrics, surpassing conventional OFDM and static AFDM. Agile-AFDM, therefore, offers a crucial step forward in the design of agile waveforms for 6G and beyond.

💡 Research Summary

The paper introduces Agile‑AFDM, a novel adaptive framework that transforms the traditionally static Affine Frequency Division Multiplexing (AFDM) waveform into a dynamic, data‑aware system suitable for 6G applications. Conventional AFDM relies on two chirp parameters, c₁ and c₂, which are preset based on worst‑case channel statistics (maximum delay and Doppler spread). This static configuration limits the waveform’s ability to exploit real‑time channel variations and the actual transmitted data.

Agile‑AFDM addresses this limitation by formulating per‑block optimization problems for the chirp parameters. The optimization is conditioned on instantaneous channel state information (CSI), the current data symbol vector, and a specific performance objective. Three representative objectives are considered:

1. Peak‑to‑Average Power Ratio (PAPR) reduction – crucial for power‑limited user equipment. The authors analytically show that PAPR is primarily governed by c₂ and exhibits a periodic structure. Leveraging this property, they propose a low‑complexity grid‑search algorithm that evaluates a discrete set of candidate c₂ values per block, selecting the one that minimizes the continuous‑time PAPR.

2. Inter‑Carrier Interference (ICI) suppression – essential for reliable communication in high‑mobility scenarios where Doppler spread destroys sub‑carrier orthogonality. The effective channel matrix in the DAFT domain, H_eff, depends jointly on c₁ and c₂. By expressing the signal‑to‑interference ratio (SIR) as a fractional function of the parameters, the paper introduces an alternating‑optimization scheme based on fractional programming. In each iteration, one parameter is fixed while the other is optimized, guaranteeing convergence to a local optimum.

3. Cramér‑Rao Lower Bound (CRLB) minimization for sensing – relevant for integrated sensing and communication (ISAC). The CRLB for delay and Doppler estimation is a non‑convex function of both chirp parameters. To navigate the complex landscape, the authors employ Particle Swarm Optimization (PSO), which efficiently explores multiple regions and converges to a near‑optimal solution.

The system model retains the conventional cyclic prefix concept, but replaces it with a chirp‑periodic prefix (CPP) that aligns with the chosen c₁. When 2Nc₁ is an integer, CPP reduces to the standard CP, ensuring backward compatibility. The transmitter maps data symbols to the time domain via the inverse DAFT, applies CPP, and sends the waveform. At the receiver, CPP removal and DAFT demodulation recover the symbols, after which standard equalization (e.g., MMSE) can be applied without modification.

Extensive Monte‑Carlo simulations are conducted over a variety of doubly‑dispersive channels, mobility levels (0–500 km/h), path counts (up to 8), and SNR values (0–30 dB). Results demonstrate:

• PAPR – Agile‑AFDM achieves roughly a 50 % reduction compared with static AFDM and a 6–7 dB improvement over OFDM.
• ICI / SIR – In high‑Doppler scenarios, the proposed alternating‑optimization yields a 7.32 dB SIR gain over OFDM and a 2.64 dB gain over static AFDM, with noticeably lower variance across channel realizations.
• CRLB – For monostatic sensing, delay‑estimation CRLB improves by an average of 8.3 dB and Doppler‑estimation CRLB by 12.7 dB relative to the baseline, especially pronounced when Doppler spreads are large.

Computational complexity is kept modest: the PAPR grid search runs in O(|𝒞₂|·N) time, the alternating‑optimization converges within a few iterations (typically <10), and PSO requires only a few dozen particles to reach satisfactory solutions. Real‑time feasibility is confirmed with average per‑block optimization latency below 1 ms on a standard CPU, and a modest increase in FPGA resource utilization when implemented in hardware.

The authors also discuss future extensions, such as employing lightweight neural networks to predict near‑optimal chirp parameters from CSI, coordinating parameter selection among multiple users to avoid collisions, and designing ASIC‑level accelerators for the optimization kernels.

In conclusion, Agile‑AFDM demonstrates that by treating the chirp parameters as controllable, per‑block variables, AFDM can simultaneously enhance power efficiency, communication reliability, and sensing accuracy—key pillars of the envisioned 6G ecosystem. The framework requires only software‑level upgrades to existing AFDM transceivers, making it a practical and impactful step toward truly adaptive waveforms for next‑generation wireless systems.