Making wireless communication more reliable in electric vehicles with a Bayesian framework
Chin-Hung Chen defended his PhD thesis at the Department of Electrical Engineering on 15 September.
The increasing use of electric vehicles (EVs) creates new challenges for wireless communication systems. For example, electronic components inside EVs generate electromagnetic interference that can be highly impulsive and bursty. Such interference differs substantially from the conventional Gaussian noise assumed in many receiver designs and can therefore significantly degrade reception quality. In his PhD research, Chin-Hung Chen developed advanced receiver techniques to improve the robustness of digital audio broadcasting (DAB)-like communication systems in the presence of EV-induced interference.
Chin-Hung Chen鈥檚 proposed approach is based on Bayesian inference and explicitly incorporates statistical and physical knowledge of both the interference and the communication system. His research begins with the statistical characterization of EV interference using measurement data. A Markov-Middleton model is used to describe its bursty time-domain behavior, while a Gaussian mixture model combined with a hidden Markov model is developed for frequency-domain processing in orthogonal frequency-division multiplexing (OFDM) systems.
New receiver algorithms
Based on these models, Chen developed new receiver algorithms for both single-carrier and OFDM systems. For single-carrier systems, he proposes an optimal receiver for differentially encoded phase-shift keying, together with an Expectation-Maximization (EM) framework that enables fully blind estimation of the communication channel and interference characteristics. A physics-aware initialization method is further introduced to improve the reliability of blind estimation.
OFDM systems
For OFDM systems, the Bayesian framework was extended to frequency-domain interference mitigation. Chen developed a block-based EM algorithm to efficiently estimate and track interference states by exploiting the null tones already present in DAB transmissions.
Overall, the research demonstrates that explicitly modeling the statistical structure of EV-induced interference can substantially improve receiver reliability compared to conventional designs based on Gaussian-noise assumptions.
Title of PhD thesis: Supervisors: Prof. Alex Alvarado and Dr. Wim van Houtum.