Research for Adversarial Planning Recognition and Reply in the Complex Domains and the More Agents Conditions

At present, the research on planning system mainly focuses on searching a valid planning in the single domain and closed environment, such as the classical planning problem, rocket problem, and briefcase problem and so on. However, in the practical problem with adversarial action planning, this kind of classical planning theory is difficult to apply to complicated and diverse world condition, such as game domain, war domain and so on. Therefore, when facing so many particular planning problems in realistic world, it should pay more attention to the valid recognition and reply to adversarial action planning between more agents in complex conditions. This paper just points to such problems. In an open world, it makes the timely and valid decomposition to adversarial planning which is uncertain, fuzzy and complicated planning sequence by constructing adversarial action model, combining the relevant knowledge and reasoning mechanism of planning recognition expert system, making use of the method of action driven to realize the process of adversarial planning recognition by matching with the basic adversarial action item in adversarial planning base. Meanwhile, on the basis of recognition process and according to the relevant knowledge in adversarial knowledge base, adopting the corresponding reply planning scheme to realize the valid precaution and counterattack to agent who executes adversarial planning. In addition, this paper also makes an outlook to the research and development of the adversarial planner.