- Clear path
- Add cost to box on path in PointerSearch
- Generalize multiagent clearPath to singleagent
- Fix all pairs shortest path preprocessing
- Merge multiagent
- Make sure they don't stand on goals (see
onGoal
) - high weighted goal heurstic is important so agent can move away from box
- Good for testing multiagent parallelization:
MABeliebers.lvl
- Heuristic based on PointerProblem (test on SAEvilCorp.lvl)
- Multiagent
- Output actions for all agent (NOP as default for agents doing nothing)
- run previous actions at the same time
- merge (oh boy)
- Goal ordering: make adjustments before solving goal greedily (see SAtest3.lvl)
- Run different goal sequence permutations in parallel
MSBuild AntStar/AntStar/AntStar.fsproj /p:Configuration=Release;Targets=Clean<Paste>
Run the project with dotnet from CLI
dotnet AntStar/AntStar/bin/Release/netcoreapp2.0/AntStar.dll [options] [file ...]
Run all test levels with script
./antstar.sh
-lvl Level file
Order by setting all goals (except goals already in "solvable") to walls expect one (removing all boxes and agents If it can be solved add to "solvable" ordered set.^M n^2^M Ordering: https://www.ijcai.org/Proceedings/89-2/Papers/013.pdf^M