Wireless innovations Next-generation
Online Workshop(WiNOW)
3-6 November, 2025 // Virtual
Mahsa Derakhshani
Loughborough University
Mahsa Derakhshani received the Ph.D. degree in electrical engineering degree from McGill University, Montr´eal, Canada, in 2013. She was an Honorary NSERC Postdoctoral Fellow with the Department of Electrical and Electronic Engineering, Imperial College London, from 2015 to 2016, a Post-Doctoral Research Fellow with the Department of Electrical and Computer Engineering, University of Toronto, Toronto, Canada, and a Research Assis tant with the Department of Electrical and Computer Engineering, McGill University, from 2013 to 2015. She is currently a Reader (Associate Professor) in digital communications with the Wolfson School of Mechanical, Electrical and Manufacturing Engineering, Loughborough University, U.K. Her research interests include machine learning and optimization for wireless communications, ultra-reliable low latency communications, edge computing, and Ambient IoT. She received several awards and fellowships, including Royal Academy of Engineering/The Leverhulme Trust Research Fellowship (2020–2021), the Natural Sciences and Engineering Research Council of Canada (NSERC) Post-Doctoral Fellow- ships (2015–2017), the Fonds de Recherche du Quebec–Nature et Technologies (FRQNT) Post-Doctoral Fellowship (2013–2015), the John Bonsall Porter Prize (2009–2010), and the McGill Engineering Doctoral Award (2008–2011). She serves as an Editor for IEEE WIRELESS COM MUNICATIONS LETTERS, IEEE Internet of Things Magazine, and IET Signal Processing journal.
Talk Title: Reinforcement Learning Approaches for Robust Satellite Communication with Outdated CSI
The integration of low earth orbit (LEO) satellites with terrestrial communication networks is a promising approach towards global connectivity. Yet, maintaining high-quality service heavily depends on having accurate channel state information (CSI). In practice, channel estimation in satellite communications is challenging due to the long propagation delays between satellites and terrestrial users, which could result in outdated CSI observations.
In this talk, I’ll explore how deep reinforcement learning approaches can help overcome these challenges and enable more robust satellite communications. We’ll look at two key directions: using the Deep Deterministic Policy Gradient (DDPG) algorithm with state augmentation techniques for precoding design of single-satellite systems, and extending the concept to multiple satellites through multi-agent reinforcement learning approaches.