In-Band Full-Duplex Antenna Systems

Abstract

In-band full-duplex (IBFD) wireless systems enable simultaneous transmission and reception in the same frequency band, offering increased spectral efficiency for emerging communication and sensing applications. Their practical realization, however, is fundamentally limited by strong self-interference (SI) from the transmit (Tx) to receive (Rx) path, which can saturate the receiver front end and degrade sensitivity. This thesis investigates antenna- and beamforming-based approaches for controlling SI while preserving desired radiation and reception performance of full-duplex arrays.An analytical framework is developed for non-shared-aperture IBFD antenna arrays using scattering parameters, power-wave representations, and embedded element patterns. Within this framework, Tx beamforming is formulated as a constrained optimization problem that maximizes desired Tx array gain while limiting SI coupled to the Rx array. The resulting eigenvalue formulation identifies a low-coupling Tx subspace whose dimensionality determines the available degrees of freedom. Receive beamforming is then employed to suppress residual SI while preserving the desired signal. Furthermore, the Tx-Rx array geometries are investigated for their impact on the coupling subspace and gain–isolation trade-off. The framework is subsequently extended to IBFD multi-user multiple-input multiple-output (MIMO) systems, combining SI-constrained Tx beamforming with zero-forcing precoding and Rx combining in the SI-channel null space to enable simultaneous multi-user operation, yielding closed-form solutions within the resulting reduced subspaces.The proposed concepts are experimentally validated using a 3x5 Tx/Rx Vivaldi antenna array integrated with a Zynq UltraScale+ RFSoC ZCU216 platform. Measurements demonstrate more than 45 dB of Tx-to-Rx isolation with constrained Tx beamforming and more than 80 dB of overall system isolation with joint Tx-Rx beamforming, while retaining the desired beamforming characteristics at 3 and 3.5 GHz. The SI-mitigation principle is further demonstrated in a full-duplex frequency-modulated continuous-wave radar operating at 3 GHz with a 100 MHz chirp bandwidth, achieving more than 40 dB reduction in coupled SI and enabling target echoes to be recovered.To complement beamforming-based mitigation, a machine-learning-assisted antenna synthesis methodology is developed for four-port full-duplex MIMO patch antennas. A symmetry-constrained convolutional neural network surrogate model combined with genetic optimization is used to explore antenna topologies with enhanced inter-port isolation, achieving measured isolation of up to 36 dB while maintaining favorable radiation performance.Overall, the thesis establishes a system-oriented approach to SI mitigation spanning antenna design, spatial beamforming, hardware implementation, and full-duplex sensing. The results demonstrate that exploiting antenna-system degrees of freedom can substantially reduce the burden on subsequent cancellation stages while maintaining the spatial performance required for full-duplex communication and sensing.