AI-Based Simulation

The objective of this research is to explore novel methods in machine learning that can enhance artificial intelligence in simulation and computer games. The eventual aim is to generate and deploy such a form of artificial intelligence in common combat or melee applications where users can be exposed to opponents/non-player entities that adaptively learn and thereby challenge their minds.

This research will focus on machine learning that is realistic, robust, and efficient for both strategic and real-time applications. We plan to investigate the usage of machine learning techniques such as, grammatical learning, neural networks, reinforcement learning, evolutionary learning, etc., for their suitability in simulation and games. In many situations, these techniques require an inconveniently large number of observations; for specific simulation scenarios, further modeling is needed to use them efficiently. This gives rise to a collection of research issues regarding the different levels of applicability of these techniques in simulation environments. Specific simulation and game scenarios that we consider for this research include combat tactics such as attack, defense and weapon/resource selection upon melee encounters. We focus on machine learning methods that can acquire, retain and apply domain knowledge. Other practical applications may include customer service software, e-learning games and embedded software in electronic devices.

Principal Investigator: Assoc Professor Narendra S. Chaudhari

Team Member:
Ms. Chen Jinmiao, Project Officer