Xiaolan Liu

Xiaolan received a B.Eng and M.Eng at Jilin University from China in 2014 and 2017, respectively, and her PhD degree at Queen Mary University of London in 2021. She is currently a visiting research fellow at King’s College London and The Hong Kong University of Science and Technology.

Her research interests are including federated learning, multi-agent reinforcement learning, edge computing and IoT networks, digital twin & Metaverse, as well as intelligent wireless communication optimization.

She has won the Best paper award from China Communications 2023.

She has the experience of working for a project Innovate UK, RCUK & MoST Newton Fund and EU Horizon 2020. She is the PI for JOINT FUND 2022 from Loughborough University London.

Professional experience

Editors

• IEEE Internet of Things Journal
• China Communications

Chairs

• Session Chair for IEEE International Conference on Telecommunications ICT2021
• Co-Chair for IEEE Wireless Communications and Networking Conference WCNC2023 - Workshop
• Symposium Co-chair for IEEE International Conference on Communication Technology ICCT2023

Publicity officer

• IEEE Next Generation Multiple Access ETI
• IEEE Women in Engineering UKI

Areas of research expertise

Adversarial attacks for distributed learning - Federated learning has been used in many subjects to train machine learning models in a distributed manner while maintaining good privacy protection. However, it is still susceptible to poisoning attacks when there are malicious users existing in the networks, which is even more challenging for VR and Metaverse applications where the generated data at each user always includes their private and sensitive information. Therefore, we first emulated the attackers’ behaviours by generating different adversarial attacks and then proposed a new type of adversarial attack scheme that induces uncertainty at adversaries based on Bayesian optimisation.

Distributed learning for heterogeneous wireless networks - The heterogeneity of deploying distributed learning in wireless networks can be summarized into three categories: 1) data heterogeneity, 2) user diversity and 3) model heterogeneity. To address user and data heterogeneity, hybrid split and federated learning architecture is proposed, also, to reduce transmission latency and save energy consumption, two important performance metrics in wireless networks, efficient user scheduling and resource allocation scheme is studied. Moreover, we also consider the asynchrony of users due to them completing local training at different times and propose asynchronous learning architectures to address this challenge. To include the contribution of the slower users at the model aggregation step, we propose performing one step of the local model update by aggregating the received global model with the stored local model updates ahead of performing local training to alleviate the effect of stale gradients computed by slower users.

Multi-agent reinforcement learning enabled resource allocation - The emerging intelligent applications at the network edge requested by multiple users naturally make the wireless network a multi-agent problem, when increasing to large-scale networks, centralized resource management inherits high communication and computation overhead. We focus on developing efficient multi-agent reinforcement learning algorithms to solve resource allocation and decision making on computation offloading with edge computing deployed in networks.

• King's College London, UK
• Queen Mary University of London, UK
• The Hong Kong University of Science and Technology
• Jilin University, China

Current PhD supervisions as a Principal Supervisor:

Haokai Yang: Personalized federated learning for Metaverse, Oct. 2023-

Current PhD supervisions as a Co-Supervisor:

Mr Rafael Moreira Pina: Scalable Multi-Agent Reinforcement Learning, Oct. 2021 –
Supervisor: Dr. Varuna De Silva

Mr Aristodemou, Marios: Secure AI on Wireless networks, Oct. 2021 –
Supervisor: Prof. Sangarapillai Lambotharan

Mr Kotagaloluwegedara Herath Mudi, Charuka Kaushalya Bandara Herath: Homomorphic Encryption and Federated Learning, Oct. 2022 –
Supervisor: Yogachandran Rahulamathavan

• IEEE Member
• IEEE Communications Society Member
• IEEE Women in Engineering Member