As part of the Artificial Intelligence and Multi-Agent Systems course offered by DTU during the A.Y. 2017/18, a mandatory final project involved the design and development of a Sokoban puzzle solver leveraging commonly known A.I. principles, both for Single and Multi-Agent systems. My groupmates and I worked together to create a solver in Java using a combination of HTN planning, optimization heuristics, and the Belief-Desire-Intention model, resulting in a fully autonomous system able to solve levels (i.e. maps) of different complexity.
The below videos show a running demo of our solver for two different levels.
Links
- Source code: Github repository