Adaptive End-to-End Transceiver Design for NextG Pilot-Free and CP-Free Wireless Systems

TL;DR

This work addresses the inefficiency of pilot-based and CP-based OFDM in NextG systems by proposing an adaptive end-to-end transceiver that operates without pilots and CP. The approach combines AI-driven constellation shaping with a neural receiver and introduces a lightweight channel adapter for efficient online adaptation, plus a scalable mechanism to support multiple modulation orders within a single model. A constrained training framework enforces PAPR targets without extra transmission overhead, and an online/offline training strategy enables rapid adaptation with minimal parameter updates. Extensive simulations on 3GPP channel models demonstrate gains in BER and throughput, along with robust adaptation and reduced model storage, highlighting practical potential for AI-native air interfaces in NextG.

Abstract

The advent of artificial intelligence (AI)-native wireless communication is fundamentally reshaping the design paradigm of next-generation (NextG) systems, where intelligent air interfaces are expected to operate adaptively and efficiently in highly dynamic environments. Conventional orthogonal frequency division multiplexing (OFDM) systems rely heavily on pilots and the cyclic prefix (CP), resulting in significant overhead and reduced spectral efficiency. To address these limitations, we propose an adaptive end-to-end (E2E) transceiver architecture tailored for pilot-free and CP-free wireless systems. The architecture combines AI-driven constellation shaping and a neural receiver through joint training. To enhance robustness against mismatched or time-varying channel conditions, we introduce a lightweight channel adapter (CA) module, which enables rapid adaptation with minimal computational overhead by updating only the CA parameters. Additionally, we present a framework that is scalable to multiple modulation orders within a unified model, significantly reducing model storage requirements. Moreover, to tackle the high peak-to-average power ratio (PAPR) inherent to OFDM, we incorporate constrained E2E training, achieving compliance with PAPR targets without additional transmission overhead. Extensive simulations demonstrate that the proposed framework delivers superior bit error rate (BER), throughput, and resilience across diverse channel scenarios, highlighting its potential for AI-native NextG.

Figures (14)

• Figure 1: Overview of conventional (pilot-aided and CP-aided) and AI-based (pilot-free and CP-free) transceiver architectures.
• Figure 2: Illustration of the proposed neural receiver architecture and the structure of the residual block.
• Figure 3: Comparative illustration of our proposed channel adapter framework versus popular and widely used alternatives. (a) depicts the typical offline method. (b) illustrates the traditional end-to-end approach with full fine-tuning. (c) Tailored for the symbol detection task, our network incorporates a lightweight adapter within the backbone to enable efficient transfer learning.
• Figure 4: Architecture of channel adapter, which employs a bottleneck structure composed of depth-wise separable convolutions and ReLU activation, followed by the AF module.
• Figure 5: AI transceiver design for supporting multiple modulation orders: (a) transmitter-side geometric constellation shaping using zero padding; (b) receiver-side LLR masking.