The 16th International Conference on Electronics, Communications and Networks (CECNet 2026)

Abstract: As data centers and B5G/6G networks face exploding bandwidth demands, single-channel optical interconnects operating beyond 400 Gb/s are emerging as a pivotal technology. In this race, IM/DD systems remain strategically indispensable for short-reach applications, offering unrivalled power and cost efficiencies over coherent alternatives. This work tackles the two core hurdles, through a multi-dimensional optimization framework that unites device engineering, advanced modulation design, and AI-native signal processing. Looking forward, we identify the convergence of probabilistic shaping, end-to-end learning-based transceiver optimization, and cutting-edge silicon photonics as the transformative triad that will unlock single-channel 400 Gb/s operation while sustaining ultra-high energy efficiency.

Abstract: Real time and high precision wireless positioning serves as a fundamental supporting technology for smart city, intelligent transportation, Internet of Vehicles, target reconnaissance, unmanned platform coordination and other electronic information and situation awareness applications. Traditional two step wireless localization first estimates intermediate parameters and then solves target positions, which suffers from information loss, performance limitation and target measurement data association ambiguity. Direct localization can obtain target coordinates straightly from raw received signals without intermediate parameter estimation, and it makes full use of signal waveform characteristics to avoid association mismatches. Nevertheless, practical non ideal scenarios bring severe challenges, including low signal to noise ratio caused by long transmission distance, non line of sight propagation, group multi targets, multi band transmitted signals, as well as array manifold and channel model errors.
This report investigates cooperative direct localization technologies coping with the above non ideal conditions. Firstly, a multi constant modulus source direct localization algorithm based on direction of arrival and Doppler shift is proposed, which exploits constant modulus signal properties to boost positioning accuracy under low SNR and small snapshot quantities. Secondly, a time difference of arrival based direct localization approach utilizing known modulation information is presented, where alternating iteration combining discrete symbol optimization and continuous position estimation improves localization performance close to the bound of fully known waveforms. Thirdly, a cooperative direct localization scheme fusing short wave and ultra short wave observations is developed. By jointly processing multi band signals, it extends effective observation baseline and optimizes geometric error ellipses for positioning. Fourthly, a calibration assisted robust direct localization method is designed to mitigate array manifold errors. Known position calibration emitters are adopted in alternating iteration optimization to eliminate estimation bias induced by array imperfections. Furthermore, a deep learning enabled over the horizon target direct localization framework based on bidirectional long short term memory (BiLSTM) network is put forward. The SL BiLSTM model extracts implicit location related features from multi station array signals to counteract complicated and uncertain ionospheric interference. Corresponding signal models, iterative optimization strategies and simulation validations are provided for each method. Simulation results demonstrate that the proposed cooperative direct localization methods achieve superior positioning accuracy compared with conventional two step and non cooperative counterparts, especially under low signal quality non ideal conditions. Finally, prospective research directions are summarized, covering group target cooperative localization under spatio temporal correlation, multi error calibration aided robust localization, bandwidth constrained cooperative direct localization, deep learning based direct tracking and group target tracking techniques.