Planning Goal-Directed Activities by an Autonomous Robot Based on Contradictory Information under Conditions of Uncertainty

Abstract

The expediency of forming and storing in the knowledge representation model of an autonomous robot contradictory information about the laws of transformation of various situations in a problem environment (PE) that occur as a result of the actions performed by the robot is substantiated. This need is due to the fact that a priori it is not possible in practice to construct and assign to an autonomous robot a detailed formal description of a model of a problem environment. The robot is actually forced to function in a priori underdetermined problem environments. This, in turn, leads to the fact that under identical conditions, according to a given model of the problem environment, but taking into account its actual characteristics, various actions performed by the robot can lead to the required result to achieve a given goal. Consequently, in real operating conditions, an autonomous robot may encounter the emergence of "contradictory" information when, under identical conditions, according to a given PS model, a formed plan of goal-directed activity, which was previously effective, requires significant adjustments to achieve a given goal. Such an adjustment to the formed behavior plan is usually associated with the robot studying the patterns of purposeful transformation of situations in the actual problem environment and replenishing procedural knowledge. Thus, the use of contradictory data associated with the incompleteness of a priori specified knowledge provides an autonomous robot with the opportunity to expand information about the patterns of an a priori underdetermined problem environment and, on this basis, increase functionality. To solve this problem, the article proposes a structure of typical elements for representing "contradictory" knowledge, including various elementary acts of behavior, the development of which allows an autonomous robot to obtain a given result by performing various actions in similar operating conditions, taking into account their individual characteristics that are not reflected in the model describing the current problematic environment situations. Cognitive tools have been developed to provide an autonomous robot with the ability to organize an effective combination of procedures for planning goal-directed behavior based on a given model of knowledge representation and self-learning procedures in a priori underdetermined conditions of an unstable problem environment. In general, the considered cognitive tools for planning the expedient activity of an autonomous robot allow to expand its functionality and adapt on this basis to complex a priori underdetermined operating conditions.