Jianhong Wang’s Academic Personal Website

Ph.D. Thesis

My Ph.D. thesis titled Shapley Value Based Multi-Agent Reinforcement Learning: Theory, Method and Its Application to Energy Network is available online.

Short Bio

I am a Senior Research Associate working with Prof Jonathan Lawry at University of Bristol, and a member of the European Lab for Learning and Intelligent Systems (ELLIS). Earlier, I was a Postdoctoral Research Associate at University of Manchester, UK, working with Prof Samuel Kaski.

I obtained my Ph.D. degree in Electrical and Electronic Engineering Research from Imperial College London, UK, in 2024, under supervision of Dr Yunjie Gu, Prof Tim C. Green and Prof Tae-Kyun Kim. Before that, I obtained my M.Res. degree in Web Science and Big Data Analytics, from University College London (UCL), UK, with distinction, in 2018, under supervision of Prof Jun Wang, as well as my M.Sc. degree in Computing (Machine Learning), from Imperial College London, UK, with merit, in 2017, under supervision of Prof Björn Schuller for the master project. Earlier, I obtained my B.Eng. degree in Computer Science and Electronic Engineering, from University of Liverpool, UK, with the first-class honours degree, in 2016, under supervision of Prof Danushka Bollegala and Prof Karl Tuyls for the final year project.

Research Interests

My primary research interests revolve around Multi-Agent Reinforcement Learning, Robust Reinforcement Learning and Ad Hoc Teamwork. About Multi-Agent Reinforcement Learning and Ad Hoc Teamwork, my emphasis lies in designing algorithms through the lens of Cooperative Game Theory. I firmly believe that integrating fundamental mechanisms from Game Theory into the development of learning algorithms can shed light on interpretability, transparency and reliability of comtemporary learning-based Multi-Agent Systems. Beyond theoretical endeavors, I am passionate about the practical applications of Machine Learning algorithms (particularly Multi-Agent Reinforcement Learning) to Autonomous Systems (e.g. Smart Grids, Robotics, Dialogue Systems, etc.).

Academic Activities

17/09/2024: Started working on the national project as a team member, EPSRC AI hubs: Information theory for distributed AI (INFORMED-AI) (EP/Y028732/1).
09/08/2024: Organizing a workshop called Coordination and Cooperation in Multi-Agent Reinforcement Learning (CoCoMARL) at the Reinforcement Learning Conference 2024 (RLC 2024). The aim of this workshop is to promote the advancement of cooperative and coordinated multi-agent reinforcement learning in both theory and real-world applications (e.g. swarm robotics, traffic management, etc.). To achieve the goal, this workshop is constituted of paper submissions, invited talks and panel discussions.
07/02/2023: Started working on the national project as a team member, Turing AI World-Leading Researcher Fellowship: Human-AI Research Teams - Steering AI in Experimental Design and Decision-Making (EP/W002973/1).

Personal Updates

25/09/2024: One paper accepted at NeurIPS 2024
17/09/2024: Joined University of Bristol as a Senior Research Associate
02/05/2024: One paper accepted at ICML 2024
13/02/2024: One paper accepted at ICAPS 2024
10/12/2023: One paper accepted at AAAI 2024
22/09/2023: One paper accepted at NeurIPS 2023
31/08/2023: One paper accepted at CoRL 2023
25/04/2023: One paper accepted at ICML 2023
07/02/2023: Joined University of Manchester as a Postdoctoral Research Associate

Selected Publications

Open Ad Hoc Teamwork with Cooperative Game Theory, accepted at ICML 2024
Robust Reinforcement Learning in Continuous Control Tasks with Uncertainty Set Regularization, accepted at CoRL 2023
SHAQ: Incorporating Shapley Value Theory into Multi-Agent Q-Learning, accepted at NeurIPS 2022
Multi-Agent Reinforcement Learning for Active Voltage Control on Power Distribution Networks, accepted at NeurIPS 2021
Modelling Hierarchical Structure between Dialogue Policy and Natural Language Generator with Option Framework for Task-Oriented Dialogue System, accepted at ICLR 2021
Shapley Q-value: A Local Reward Approach to Solve Global Reward Games, accepted at AAAI 2020 (Oral)

Open-Source Environment

MAPDN: An environment of distributed/decentralised active voltage control on power distribution networks and a batch of state-of-the-art multi-agent actor-critic algorithms that can be used for training.