Logo image
Back
Meta-Reinforcement Learning Optimization for Movable Antenna-aided Full-Duplex CF-DFRC Systems with Carrier Frequency Offset
Journal article   Peer reviewed

Meta-Reinforcement Learning Optimization for Movable Antenna-aided Full-Duplex CF-DFRC Systems with Carrier Frequency Offset

Yue Xiu, Wanting Lyu, You Li, Ran Yang, Phee Lep Yeoh, Wei Zhang, Guangyi Liu and Ning Wei
IEEE Transactions on Communications, Vol.Advanced access
02-Mar-2026
DOI: https://doi.org/10.1109/TCOMM.2026.3669488

Has preprint

Abstract

cell-free carrier frequency offset dual-functinal radar communication manifold optimization penalty dual decomposition meta-reinforcement learning
By enabling spectrum sharing between radar and communication operations, the cell-free dual-functional radar–communication (CF-DFRC) system is a promising candidate to significantly improve spectrum efficiency in future sixth-generation (6G) wireless networks. However, in wideband scenarios, synchronization errors caused by carrier frequency offset (CFO) can severely reduce both communication capacity and sensing accuracy, especially when multiple geographically distributed full-duplex (FD) access points (APs) are jointly coordinated. In this paper, we consider a wideband FD CF-DFRC system where each AP is equipped with movable antennas (MAs). This setting is fundamentally different from existing DFRC or MA-aided designs that typically assume fixed-position antennas, half-duplex operation, or perfect synchronization. First, we develop a field-response-based channel model and derive a worst-case weighted communication–sensing rate (WCSR) that explicitly captures the impact of inter-AP CFO on both the uplink communication signal-to-interference-plus-noise ratio (SINR) and the radar echo SINR. Our analysis reveals that CFO increases the Cramér–Rao lower bound (CRLB) of target position estimation, thereby degrading sensing accuracy. Based on this characterization, we formulate a robust worst-case WCSR maximization problem that jointly optimizes MA positions, transmit beamforming vectors, receive filters, and CFO-related parameters under transmit power and MA-position constraints. To tackle the resulting highly non-convex problem, we propose a two-stage robust optimization framework. In the first stage, we employ fractional programming together with manifold optimization (MO) and penalty dual decomposition (PDD) to solve the worst-case CFO subproblem on the complex unit-modulus manifold, thus obtaining a CFO-robust closed-form structure for the WCSR. In the second stage, we design a meta–reinforcement learning (MRL) based resource allocation scheme that jointly optimizes the MA positions and beamforming vectors in a data-driven manner for dynamic wireless environments. Unlike conventional deep reinforcement learning (DRL) methods, the proposed MRL framework learns a meta-policy that can rapidly adapt to varying channel and CFO realizations, substantially improving convergence speed and scalability. Simulation results show that the proposed robust MO–PDD–MRL framework significantly outperforms existing DRL-based and non-robust CF-DFRC schemes in terms of both communication and sensing performance under CFO impairments. Furthermore, compared to fixed-position antenna (FPA) architectures, the MA-aided CF-DFRC system exhibits markedly enhanced robustness and adaptability to CFO effects and target mobility.

Details

Metrics

1 Record Views
Logo image