This thesis proposes a quantum machine learning (QML) technique for demapping irregular circular QAM modulations for 5.5G and beyond. Future modulation trends are going beyond 1024-QAM, especially for Wireless-Fidelity (Wi-Fi), whose complexity is toward to exponential. Machine learning offers the potential to accelerate the learning of detection patterns from received signals, while quantum superposition enables parallel computation. The integration of these two approaches presents a promising strategy for handling irregular modulation schemes.
This thesis uses amplitude encoding technique to decode the input dataset or transmitted information symbol. Then, a Hadamard gate is used to create qubits in a superposition state. This thesis utilizes the empty states to observe the probabilities of the designed qubits for a given signal. As the final step, this thesis determines the bit based on the smallest probability obtained from the amplitude of the quantum states.
The results of this thesis show that the proposed QML-based technique for demapping requires fewer number of qubits compared to the existing QML and has lower estimated execution time compared to the classical demapping for a higher constellation number. We successfully detected complex binary phase-shift keying (C-BPSK) modulation using only three qubits and the addition of one qubit for a higher constellation number. This thesis also proposes irregular circular modulation for the 64, 1024, 2048, 4096, and 8192 QAM constellations which are detected with QML-based demapping technique. The proposed technique is expected to be applied to beyond fifth-generation (5G) technologies that require high data rates and low latency. The results of this thesis are also expected to be useful for the development of sixth generation (6G) technology.