NOTE: This is the latest in a series of Q&A stories featuring Brock University faculty members who are integrating the Niagara 2022 Canada Summer Games into their research projects. For more information on Brock’s academic activities around the Games, visit brocku.ca/canada-games
Assistant Professor of Computer Science and Biological Sciences Yifeng Li’s research interests include artificial intelligence algorithms, machine learning, theories and applications of big data analytics, structured and unstructured data modelling, heterogeneous data integration, data mining models to bioinformatics and health informatics, among other topics.
Li is one of eight Brock researchers and scholars who received funding under the 2021-22 round of the VPR Canada Games Grant program. Here, he discusses his research project titled “Sport AI Enabled by Multi-Agent Deep Reinforcement Learning Approaches.”
Please give a brief overview of your research project
In this research, we are interested in developing artificial intelligence (AI)-driven models that can learn from experiences and gain near human-level intelligence in sport games that have many players. The research will involve several AI agents. An agent is an entity — such as AI assistants Alexa and Siri — that perceives its environment and acts autonomously to achieve specific goals. In computer games, the agent is a computer player.
To pursue our objectives, we will be exploring reinforcement learning algorithms. Reinforcement learning trains computer agents using experiences collected from interactions with the environment to enable agents to make the best or near-best decision in a particular situation. We will investigate and develop these algorithms for soccer games involving five to 11 players on each team. The performance and behaviours of AI agents will be analyzed to identify novel winning strategies.
What do you expect will be the outcome of your research?
We expect to discover novel winning tactics through the developed AI algorithms. These tactics can be used as inspirations for game team training. The developed AI algorithms will be deployed as computer software useful to the sport community.
How will this contribute to knowledge or understanding of the Canada Summer Games?
The novel winning tactics we discover can aid in team training as AI agents can search for, and identify, sequences of decisions that lead to a win but were never thought about by humans. Furthermore, this project will promote awareness of AI techniques in advanced game development and the combination of AI with sports. In our simulation software, if Brock and Niagara elements are embedded in the background, it would achieve a brand effect.
How did you become interested in this research?
I developed a new AI course in Computer Science — COSC 4/5P83 Reinforcement Learning. Reinforcement learning is one area in AI for sequential decision-making. In recent years, it has been widely applied to train computer agents to gain intelligence in various video games and robotics. Inspired by this, I believe reinforcement learning is a highly potential AI technique for sport games as sport games are naturally sequential decision-making processes.
How do you plan on sharing your research?
The developed tools will be shared with local athletes. We plan to report our research findings through peer-reviewed publications and share our implementations on open-source repositories such that the impact and benefit of our research can be optimized.
Do you have any advice or tips on how colleagues in your Faculty can incorporate the Canada Games into their research?
Innovation has no border. Novelty comes from bold and ambitious thinking. Take actions as an athlete if you have a goal.