Experimental model of a multi-robot system for full-scale testing of automatic blockage analysis and removal tasks is developed. The system is composed of autonomous with the appropriate set of software for planning movements of the mobile platform and the on-board manipulator using visual feedback. We describe software and algorithmic tools for scheduling and distributing subtasks, as well as coordinating movements of the robots. Software for the strategic level of group control allows determining the stages of operations for grasping and transporting individual elements of the blockage, which were detected during preliminary exploration. A knowledge base on the stages and specifics of technological operations allows for correct processing of information and making appropriate decisions in the process of multi-robot system operation. Proposed action planning algorithms are built on the basis of the finite state machine theory. Experimental studies have proved the reality of the prospects for creating multi-agent systems based on autonomous robots for solving complex problems in conditions of uncertainty.