Planning the Behavior of an Autonomous Flying Robot in a Subtask Space. Inference Procedures

Abstract This article describes a model of the representation of knowledge of autonomous flying robots in the form of a set of typical domain-independent subtasks that make it possible to construct complex programs of goal-directed activity under a priori uncertainty. The solution of the test problem is presented, which shows the effectiveness of the proposed model for the construction of an intelligent problem solver for various autonomous agents. Procedures of knowledge processing are developed, which allow flying robots to plan goal-directed behavior on the basis of automatic growth of the reduction network model for solving complex problems in the space of subtasks. Boundary estimates of the complexity of logical inference procedures are found, which confirm that the proposed knowledge representation model allows flying robots to automatically plan goal-oriented behavior with polynomial complexity under uncertainty.