BDI. There is a common goal (top-down). To solve it, nodes can request tasks. Since task will be limited to colors, the tasks will be assigned in a directed or limited manner (directed contract and/or limited broadcasting).
- Problem recognition. Check state solution, subdivide into problems (use colors and heuristics).
- Deadlock. Solve when agent is blocked. Agents are blind to boxes/agents that are not their color to find a path. Broadcast new task to unblock and bidding by heuristics.
- Task broadcasting. Request someone to do the task.
- Biding. Compare h.
- Solve problem.
The common goal is subdivided and assigned to agents (bottom-up). The smaller subproblems will be solved independently and combined.
- Problem subdivision.
- Sub-problem assignment.
- Solve each subproblem.
- Combine agents and subproblems.
- Repeat 3-4 until solution is found for top problem.
- Color.
- Initial Problem recognition (subdivide goal state) top manager.
- Blind path solving.
- Bidding.
- Broadcast new tasks (agent-manager) in the way.
- Task costs.