Problem

Conventional channel modeling approaches (empirical, deterministic, semi-deterministic) struggle to balance simplification and accuracy in complex, dynamic wireless environments. The emerging space-air-ground-sea integrated network (SAGSIN), with its diverse services, extended-spectrum operations, high-mobility scenarios, and multi-band cooperation, places unprecedented demands on highly accurate and adaptable channel models. Traditional channel sounding methods are limited by generalized measurement campaigns, inadequate cross-band consistency, high costs, limited coverage, and insufficient real-time adaptability, making them unsuitable for the fine-grained, scenario-specific, and high-precision channel modeling required by SAGSIN.

Method

The Integrated Channel Sounding and Communication (ICSC) framework proposes a unified approach that combines wireless communication and sounding functionalities. It utilizes existing communication signals for simultaneous information transmission and channel sounding, extracting inherent features like multipath delay, Doppler shift, and angle information during data demodulation. This enables real-time, continuous, and cost-effective acquisition of dynamic channel characteristics across diverse scenarios (space, air, ground, sea) and multiple frequency bands. The ICSC system actively acquires real-time Channel State Information (CSI), IQ samples, and other channel features. Leveraging AI algorithms such as Convolutional Neural Networks (CNNs) for intelligent scenario identification (SI) and Reinforcement Learning (RL), specifically Dueling Double Deep Q-Network (D3QN), for dynamic waveform optimization, ICSC adapts waveform parameters (e.g., modulation scheme, coding rate) in real-time. This adaptive mechanism is fed back to the transmitter, ensuring optimal communication strategies, enhancing performance, and reducing system overhead. An Integrated Verification System (IVS) was developed using USRPs, LabVIEW NXG, and IEEE 802.11ac standards at 5.9 GHz. It incorporated a CNN-based SI module (5 convolutional layers, 32 filters, 9x9 kernel) and a D3QN-based decision module (64-dimensional feature extraction, separate value and advantage branches) to select optimal Modulation and Coding Schemes (MCS) based on SI and Signal-to-Noise Ratio (SNR).

Results

The IVS successfully demonstrated the feasibility and effectiveness of the ICSC technology. Experimental results showed that the system accurately measures and captures CSI, reliably identifying communication scenarios (EPA, TDL-C, TDL-E) with over 98% accuracy in 500 SI trials. The ICSC-based waveform decision method efficiently selected and rapidly adapted optimal waveform parameters to changing scenarios. Compared to conventional SNR-based adaptive coding algorithms, the ICSC approach achieved at least a 10% improvement in average throughput without increasing the average bit error rate (BER). The IVS provided real-time visualization of communication and measurement data, confirming its ability to adapt and maintain stable communication performance in dynamic environments.

Implications

The ICSC framework offers significant potential applications. It facilitates the construction of Channel Knowledge Maps (CKMs) by providing real-time, high-precision CSI and optimized waveform parameters, addressing limitations of conventional CKM approaches. ICSC-enabled AI channel modeling, leveraging continuous high-resolution spatio-temporal-frequency channel data, can promote deeper analytical insights, more accurate predictions, and better cross-scenario generalization, overcoming limitations of traditional empirical models. Furthermore, ICSC provides crucial channel data support for Environmental Intelligent Communication (EIC), enabling adaptive air interface configurations and improved system reliability. Future research directions include the design of next-generation waveforms (e.g., Affine Frequency Division Multiplexing - AFDM) for enhanced resilience to doubly selective channels, joint estimation and compensation of timing offset (TO), carrier-frequency offset (CFO), and multipath fading under real-time constraints to improve CSI accuracy, and multimodal channel modeling that integrates antenna parameters, environmental context, terminal states, and device-related factors for a more holistic and accurate channel representation.