INTELLIGENT STRATEGY PLANNING AND CONTROL OF A GROUP OF MOBILE ROBOTS UNDER CONDITIONS OF INCOMPLETE INFORMATION

Abstract

Different problems of strategy planning and control of a mobile robot group under complex dynamic conditions with incomplete information about the external environment are considered. Approaches to solving problems of effective work scheduling under conditions of inconstant active group composition, searching for the source of a nonstationary concentration field, supervisory control of discrete-event systems are presented. An original mathematical model formulated in terms of work-shift scheduling problems and a problem-oriented modification of evolutionary algorithms with a specialized set of heuristics for its efficient solution are developed for the problem of scheduling top-level group work. Searching and monitoring the source of the nonstationary concentration field is carried out using a decentralized multi-agent control strategy that combines elements of bionic and gradient approaches, as well as a method for generating artificial potential fields. The considered control strategy has low computational complexity, high variability with respect to the types of fields surveyed, and is easily scalable to control any available number of mobile robots. The latter is of special importance, in particular when considering the problem of parallel and independent monitoring of multiple sources. It is proposed to use the means of logical inference, namely automatic theorem proving in the calculus of positively constructed formulas, to solve various problems of the supervised control theory of discrete-event systems used at different levels of the robotic complex hierarchical control system. Features of the calculus allows solving complex problems of dynamic systems control, as well as processing and controlling events based on environmental data in real time in the process of logical inference efficiently. The approach based on positively constructed formulas allows studying the properties of automata-based discrete-event systems, as well as to synthesize and model finite automata for the construction and realization of monolithic and modular supervisors. A general scheme combining the considered approaches for controlling a group of mobile robots at different levels and time scales within a single hierarchical control system is proposed.