In this paper, we report a continuing research work which is pursuing the further development of recipe structures, which are plan structures designed to facilitate the passing of information and knowledge among autonomous and heterogeneous agents existing in a hostile dynamic world. In particular, we present an intelligent agent architecture and study the behaviour of an agent executing a recipe in a multiagent hostile environment, in this case, a hostile fire domain. The domain and the agents raise a number of challenges for planning and plan execution which range from plan failure, real time constraints to continuous planning, and more notably inter-agent co-operation and communication. The concept of the recipe has proved to be very promising in our simulated fire world.
展开▼