Learning and Optimization in Games

Game AI Optimisation AI Ethics Scenario Generation

F unctional U seful N ovel

Research Interests

For more about our research, please check our publications or our PI's Google Scholar.

AI in/for Games

Procedural Content Generation (PCG): Leveraging deep learning, reinforcement learning (RL), evolutionary computation (EC) and large language models (LLMs) to automatically and adaptively create novel and diverse video game levels, rules, and scenarios.
AI for Game-Playing: Developing advanced algorithms (particularly EC-based and RL-based) for autonomous game agents: general game playing (GGP), general video game playing (GVGP), autonomous racing.
AI in Education & Educational Games: Researching methods to turn digital games into interactive tools that maximise learning efficiency and revolutionize CS/AI education.

Trustworthy AI

Algorithmic Fairness & Robustness: Developing trustworthy AI frameworks that guarantee fair downstream decision-making processes and outcomes in stationary or non-stationary environments characterised by data imbalance, concept drift, and environmental uncertainty.
Ethical Concerns in LLMs: Investigating hidden biases/discrimination when LLMs are deployed in multiplayer environments like the game Werewolf (Mafia).

AI for Smart Cities

Vehicle Routing & Material Handling: Developing search-based and neural solvers for vehicle routing variants in real-world smart logistics.

Dynamic Material Handling & Automated Guided Vehicles: Developing (evolutionary RL) solvers for dynamic scheduling in real-world manufacturing scenarios.

Autonomous Driving: Generating controllable, multimodal driving scenarios and motion patterns to rigorously stress-test autonomous driving systems and virtual racing simulators.

Learn to Optimise under Uncertainties

Lifelong Adaptive Operator Selection: Continually learning which search operators to apply, and when, from accumulated optimisation experience across problem streams, so that the choice adapts online without retraining from scratch.

Algorithm Portfolios: Automating the configuration and selection of black-box optimisers (Meta-Black-Box Optimisation), including LLM-assisted algorithm design, to reduce manual engineering effort.

Dynamic Multi-Objective Optimisation: Researching how evolutionary algorithms can adapt to changing environments while balancing multiple conflicting objectives, and improve generalisation and robustness.

Paper accepted at ICLR 2026

2026-06-02

The paper "Lifelong Learning with Behavior Consolidation for Vehicle Routing" by Jiyuan Pei (裴季源), Yi Mei, Jialin Liu, Mengjie Zhang, and Xin Yao has been accepted at the International Conference on Learning Representations (ICLR) 2026. Congratulations!

Journal paper selected as cover paper

2026-06-02

The paper "AutoSceCraft: Generate Various Driving Scenarios from Scratch for Autonomous Driving Systems" by Wenxing Lan (蓝文兴), Jialin Liu, Bo Yuan, and Xin Yao, published in Tsinghua Science and Technology (vol. 31, no. 2, pp. 1282–1305, 2026), has been selected as the cover paper of the issue. Congratulations!

Three tutorials at IEEE WCCI 2026

2026-05-27

Our group co-organises three tutorials at IEEE WCCI 2026:

Computational Intelligence for Games (CI4Games) — by Jialin Liu (Lingnan University), Julian Togelius (New York University), and Georgios Yannakakis (University of Malta). Covers computational intelligence techniques for learning, planning, and designing in board games and video games, including evolutionary reinforcement learning, rolling horizon planning, procedural content generation, and recent advances of LLMs in games.
Computational Intelligence Approaches to AI Ethics and Governance — by Changwu Huang (SUSTech), Jialin Liu (Lingnan University), Jim Tørresen (University of Oslo), and Xin Yao (Lingnan University). Covers fairness in machine learning, explainable AI, multi-objective approaches to FairML, AI transparency, and ethical challenges in embedded systems and robotics.
Evolutionary Computation Approaches to Capacitated Arc Routing Problems — by Hao Tong (Lingnan University), Jialin Liu (Lingnan University), and Xin Yao (Lingnan University). Covers CARP variants, EC-based optimization algorithms, theoretical analyses, benchmark problems, and simulation platforms for capacitated arc routing.

JOIN US

We have multiple positions available at the levels of PhD students (admission in January/September 2027), Postdoctoral Fellows and Research Assistant at Lingnan University (Hong Kong SAR).

If you are interested, please send us your detailed CV (jialin.liu_AT_ln.edu.hk).

If you are interested in doing a PhD with our group at Lingnan University (Hong Kong SAR), there are two ways (scholarships). Note that the application submission deadline is 1 December 2026 for the Hong Kong PhD Fellowship Scheme and mid-January 2027 for admission to MPhil/PhD programmes directly.

jialin.liu_AT_ln.edu.hk