Traditional mathematical models for communication-system design often use simplifying assumptions and are difficult to extend to complex or rapidly changing channels. In recent years, deep learning, and in particular autoencoders (AEs), has enabled end-to-end optimization of transmitters and receivers and has shown strong performance in many communication scenarios. This paper provides a comprehensive review of AE-based communication systems across wireless, optical (fiber and wireless), semantic, and quantum domains. The review is based on recent works indexed in major digital libraries and is organized around four cross-cutting design themes: channel modeling and differentiability, complexity and scalability, generalization and model mismatch, and data scarcity and realism. For each theme, as well as for each application domain, we summarize representative AE architectures, training strategies, reported performance, and practical limitations. The paper also outlines how to quantify the computational complexity of AE-based transceivers using Big-O notation and discusses implications for deployment in resource-constrained environments. Finally, we identify open challenges, including robust learning over non-differentiable or partially known channels, scalable and hardware-aware architectures, and standardized evaluation protocols, and we provide guidelines for future research on AE-based designs for next-generation communication systems.
21 July 2025 in IEEE Transactions on Communications
This paper presents cyclic shift phase rotation (CSPR), an innovative technique designed to mitigate the peak-to-average power ratio (PAPR) in coherent optical orthogonal frequency division multiplexing (CO-OFDM) systems. CSPR demonstrates notable improvements when compared to traditional techniques, such as discrete Fourier transform spread (DFT-spread) and selective mapping (SLM). CSPR eliminates the need for side information, additional inverse fast Fourier transforms (IFFTs), and complex search processes that are needed by other techniques. Analytical evaluations confirm that CSPR requires O(N log C0) operations, which is significantly lower than the O(M · N logN) complexity of SLM and the O(L · n log n) complexity of DFT-spread. The CSPR has been evaluated within the CO-OFDM transmission. At a distance of 800 km, CSPR facilitates a 3 dB reduction in the required optical signal-to-noise ratio (OSNR) necessary for achieving hard-decision forward error correction (H-FEC) when compared to DFT-spread utilizing 112 sub-bands, and it exceeds the performance of the SLM at low OSNR with the use of 64 candidates. Moreover, it exhibits an 8 dB enhancement in the Q-factor over transmission distances extending up to 2000 km under nonlinear conditions. These results position CSPR as an effective and efficient solution for PAPR mitigation, significantly improving system performance, and reducing hardware complexity.
The rapid growth of the electromobility field, driven by the increasing demand for electric vehicles as a substitute for combustion engine cars, is primarily attributed to continuous optimization and improved battery and motor performance. The digital twin technology has recently emerged as a valuable tool in the electromobility field, facilitating the implementation of diagnostic and prognostic algorithms. Digital twin technology provides various benefits, including increased vehicle lifespan and optimal performance, ensured health monitoring and maintenance, as well as optimized battery management and efficiency. In this context, the paper reviews various models utilized for health monitoring, predictive maintenance, and energy consumption in electric vehicles as part of developing the E-Mule digital twin.
Polarization mode dispersion in optical fibers corrupts polarization-encoded quantum states through random, time- and frequency-varying rotations. Because quantum states cannot be directly measured and reused without disturbance, practical mitigation requires adaptive compensation that operates on a physically consistent state representation. This paper models the quantum polarization channel and compensator directly in the quaternion domain, where unit quaternions naturally represent SU(2) polarization rotations and preserve the unit-norm constraint of valid quantum states. Building on this model, we introduce a fading-memory quaternion multiplicative extended Kalman filter, which is a quaternion formulation augmented with a fading (forgetting) mechanism that selectively discounts outdated information so the estimator can rapidly re-track varying rotation and recover quickly after abrupt polarization changes. A comparative study against quaternion least mean squares, quaternion recursive least squares, and a baseline quaternion multiplicative extended Kalman filter shows that Kalman-based approaches achieve faster convergence and higher broadband spectral fidelity, while the proposed fading-memory multiplicative extended Kalman filter provides the strongest robustness in dynamic conditions by improving re-lock behavior and reducing outage events under stronger drift. These results indicate that quaternion-based channel modeling combined with fading-memory Kalman estimation is a strong approach for quantum-state-preserving polarization mode dispersion compensation in realistic fiber links.