Our PlanMerger tries to find effective and cost efficient solutions for collisions in Multi-Agent-Pathfinding. To improve scalablity and performance the problem instance with multiple agents will be divided into multiple instances with one agent each. These instances are solved by a solver separately and subsequently merged into the final solution.
As basis and benchmark framework we use asprilo. Asprilo supports many benchmark scenarios, but for the sake of simplicity we will focus on the m-domain of the instances, so only on the movement of the agents/robots. Further implementations will be done with Python or Anser Set Programming, specifically clingo.
- './encodings/m/' contains all encodings for the m-domain
- './encodings/merger/' contains different encoding versions for the plan merger
- './instances/benchmarks/' contains benchmark instances which are used for testing the plan merger encodings
- Tom Schmidt tom.schmidt@uni-potsdam.de
- Hannes Weichelt hannes.weichelt@uni-potsdam.de
- Julian Bruns julian.bruns@uni-potsdam.de