The major purpose of our research is to study cooperative planning for multi-agent system in time-critical environment. The RoboCup simulator league is the most interesting target for our research. Each agent of the team 11monkeys is realized as an autonomous agent that has two sets of rules; namely state dependent rules and position specified rules. We call the former dynamic role and the latter static role; since it is assigned to agents before the game begins and remains as they are until drastic formation change occurs. On the other hand; state dependent rules are dynamically changing according to the state of the agent; for example it has a ball or is near the ball. When the agent changes its dynamic role; robustness is the most important issue. Since our agents are autonomous; dynamic role change are triggered by their own recognition of the current state. For cooperative planning; we propose three layers planning; strategy; group; and individual.