Whilst at The University of Edinburgh I self-proposed my final year project on Plan Recognition in R.I.S.K. I was supervised by Dr. Michael Rovatsos (an awesome supervisor I might add!). The research I did was fortunately noticed by the great folks over at the CREST Center UCL in their BSc Final Year Computer Science Project Competition 2013, and it had me shortlisted as a finalist.
The project involved the design and implementation of a plan recognition agent for the board game R.I.S.K. The implementation was done in a game called Domination, a Java version of the classic board game R.I.S.K. developed by Yura.net. Domination is freely available and you can download it through Sourceforge here.
The agent that was created is based on Geib and Goldman’s Probabilistic Hostile Agent Task Tracker (PHATT). Acting as an observer to the game it’s goal is to infer the unknown missions cards of players based their actions in real time. In my project report I evaluated the plan prediction accuracy of the agent with four experiments, the results of which I presented and analysed. I then went on to discuss the project outcomes and future work.