Return to Article Details Multi-agent Learning by Trial and Error for Resource Leveling during Multi-Project (Re)scheduling