For a Research Internship in the first half of the second year of the Master AI program, I implemented a Hierarchical Task Network (HTN) Planner for the Unreal Engine 4 (UE4) game engine. A novel approach for rapid (real-time) re-planning was implemented, which makes use of previously generated plans to more quickly find new plans. The basic idea is that an NPC in a video game may often want to generate new plans (for example due to new observations which were not taken into account when generating the previous plan), but such a new plan will often not deviate very much from the previous plan. The old plan can then still provide useful information for the re-planning process.

This research resulted in the following conference paper: Dennis J.N.J. Soemers and Mark H.M. Winands (2016). “Hierarchical Task Network Plan Reuse for Video Games”. In 2016 IEEE Conference on Computational intelligence and Games (CIG 2016), pp. 1-8. IEEE.

A more detailed report is available on github, as well as the source code.