Methods of synthesis of intelligent information technology for monitoring, forecasting and resource management and reconfiguration of multi-functional groups of ground and space dynamic objects
- The development of an object-oriented specification has been completed, which has become the basis for the four subsystems of a geographically distributed simulation-modeling stand: an experimental sample of a software module (ES SM) for planning and implementing the strategies of the internal and external reconfiguration of a multifunctional grouping of dynamic ground-based and space-based objects (MGDGSO), ES SM for creating user interface, ES SM for asynchronous message exchange, ES of a complex of semi-full-scale models of ground-based and space-based robots. The use of a unified specification in heterogeneous subsystems enabled us to reach constructive consistency between the concepts of logical-dynamic models and the concepts of the problem-oriented robot learning language. The study of the ES SM performance has shown a high level of system fault tolerance due to multi-stage system integrity monitoring. Monitoring of the ES SM state is performed by the operating system services (systemd). The RTС control is carried out by the specific ES SM based on the data of generalized telemetry. The interaction of the subsystems is controlled at the application level in ES SM (compensating logic of the module operation), at the presentation, session and transport levels in the service bus (compensating actions in BPMN processes), at the lower levels – by network services.
- The service bus developed by the implementers of this project has undergone a major upgrade in terms of work on simple SNMP and MQTT protocols with autonomous controllers with sensors and actuators installed in robotic complexes. At the same time, the ES SM for service registration and request routing has been upgraded and transferred from the fixed service addresses to the Indirect Endpoints technology. To reduce the high risk of failure to perform a given work scenario, fault tolerant service connection points (Failover Endpoints) have been added. The main point of the modernization is to create several instances of each service and combine them into a failover group. The service bus automatically tracks operations’ execution and, in the event of a failure, redirects the request to the next connection point in the group. Thus, when operating in a geographically distributed heterogeneous environment with an unstable communication channel, the reliability of the developed simulation-modeling stand increases significantly, unlike the traditional openSAFETY or POWERLINK protocols, which focus on the speed of interaction in broadband networks.
- The calculations of the plans for the RTC operation with the use of optimization and heuristic algorithms have shown the following results: the effect of the use of optimization algorithms increases with an increase in the number of conflicts between operations for a resource, that is, situations in which it is necessary to make a reasonable choice of an operation from a list of acceptable ones. With two technological processes (low conflict rate), the generalized indicator of the quality of the synthesized work plan improved by 16.2%. With three processes, the improvement in the generalized quality indicator reached 27.4%. A similar solution for four processes gave an improvement of 35.4%. With a further increase in the number of production processes, the result fluctuates slightly, but remains at a level of 30–38%. The obtained results confirm that it is reasonable to apply the optimization algorithms in a technological process with a high degree of conflict between operations.
- A method has been developed for multi-criteria evaluation of the effectiveness of intelligent information technologies (IIT) for dynamic control of MGDGSO in a complicated environment and the choice of the most preferable technologies based on the proposed original complex of deterministic, stochastic and interval logical-dynamic models of MGDGSO proactive control. The methodology allows, first of all, to perform (on a unified methodological and model-algorithmic basis) stating and solving the tasks of complex modeling of MGDGSO that function in a dynamically changing environment, as well as the tasks of simultaneous synthesis of technologies and programs (plans) for managing operations, flows, resources of MGDGSO, and also, correcting (re-planning) these programs. And, secondly, it allows to solve the tasks of assessing the information technology capabilities, robustness and sustainability of certain plans when analyzing their feasibility. The method itself includes two main phases (stages): the first phase should be the formation (generation) of acceptable variants of the multi-structural macro-states of MGDGSO or, in other words, a structural-functional synthesis of a new MGDGSO state should be carried out in accordance with to the emerging (forecasted) situation. For the considered class of multi-criteria evaluation of the efficiency of using IIT, as a result of the implementation of the first phase, a Pareto set (a set of nondominated multi-structured macrostates) is formed in the space of multi-structural macrostates, ensuring that tasks assigned to the MGDGSO are performed at a fixed time interval. In the second phase, the tasks of choosing the most preferable program for proactive control of the MGDGSO structure dynamics are solved, the trajectory of which in the multidimensional phase space falls into one of the points of the Pareto set of multi-structural macrostates formed in the first phase, thereby ensuring the solution of the initial task of multi-criteria estimation and selection of the most preferable technologies and programs of proactive MGDGSO control and most relevant IIT, ensuring their successful implementation. The results of the research show that this method can be, after minor modifications, used for developing recommendations on reconfiguring software-mathematical and information software structures, reconfigurable MGDGSO control systems in a dynamically changing environment (including the occurrence of computational and off-design abnormal situations).
- The modification and generalization of the previously developed method for the selection of auxiliary criteria for the MGDGSO control with the help of reinforcement training has been performed. The obtained version of the method, which is able to control the optimization with the use of several local search heuristics, was applied to solve the task of MGDGSO routing with consideration of time intervals, having five target and seven auxiliary optimization criteria. The results of experimental studies have shown that finding the best solutions to this task necessarily requires the use of a multi-stage optimization process involving more than one heuristic and more than one criterion. The statistically significant advantage of using the method of transferring learning outcomes, carried out by using the averaged Q-values obtained in previous runs of the algorithm, as initial values for the next run, has also been revealed.
- A correction has been made to the composition and structure of atypical operators of the control language of the robot, the basic part of which determines the law of the robot movement (it is formed in the learning process), and also, the correction of their functions has been carried out by including a large class of “power operators” as part of the operators. They determine the force of the interaction of the working tool of the robot with objects, the possible movements of which are limited to bare links. This dramatically expands the functionality of remote-controlled space robots, allowing them to be used to perform assembly operations, which are currently most in demand when working in space. The correction of the structure and composition of typical operators of the robot control language was made by including conditional and unconditional jump operators, loop operators and other similar operators not affecting the movements of the robot as part of the language of the operators, as well as the inclusion of operators specifying the position of the robot in the coordinate system of its body, which allows you to successfully perform required operations regardless of possible changes in the external environment, as well as the inaccuracies of the environment model.
- It has been experimentally confirmed that it is possible to teach a robot by showing working operations with free movements of the head and arm of a human operator with reliable control of the results of learning based on a framestructured model of the form of movements (MFM). The description language used in the MFM is a multi-level, hierarchical frame system containing a semantic description of form elements and topological (metric) characteristics, as well as methods and procedures typical for various types of movements and behavior scenarios. Algorithms for analyzing the natural movements of the hand and the head of a human operator (without using traditional skeletons) have been developed, which allow to automatically obtain semantic descriptions of the form of movements in the form of a multilevel MFM frame system. Teaching algorithms by the method of showing working operations for a group of remote-controlled assistant robots, including using the 6-axis control handle, have been developed and experimentally tested. Methods and algorithms for group training and telecontrol of work operations and scenarios of behavior by robots, transfer of functions in a group, and spatial reconfiguration of a group have been experimentally investigated. The effectiveness of the developed training algorithms and the correction of scenarios and working operations when telecontroling MGDGSO in terms of interference and delays of Internet signals has been experimentally verified.
- A computer study of the processes and evaluation of accuracy indicators (characteristics) in force torque control systems when a robot performs assembly and other contact operations has been performed. The computer study of force torque control systems has been performed with the use of the MATLAB package tools. This environment was selected for the study due to the graphic tools for high-quality animation of robot and master mechanism (handle) movements. With the help of the functions of the Robotics Toolbox library in the MATLAB package, a program has been developed for creating animated models of a handle and a robot for semi-automatic position and force torque control systems. A special program file has been created (“Sim Open and Closed Force Control Systems with Torque Drives”). Three ways of moving the handle have been considered. The current study allows us to obtain an estimate of the accuracy of a human operator – by visual comparison of the programmed and real positions of the robot and the handle and by creating graphs to obtain accurate error values.
- A geographically-distributed simulation-modeling stand consisting of three main parts has been developed: an experimental robotic complex, which includes an adaptive local correction system for the trajectories of robots; experimental mathematical software of the simulation-modeling complex, including modules for solving the tasks of monitoring the state of MGDGSO, generating navigation and ballistic data, and analytic and simulation modeling for multicriteria evaluation of the effectiveness of using existing and prospective IT for MGDGSO control, planning and implementing the strategies for internal and external reconfiguration of MGDGSO; an experimental robotic complex that includes remotecontrolled assistant robots trained by an operator using the movement display method, as well as a virtual (computer) robot.
The project is implemented with financial support of Russian Science Foundation, grant № 16-19-00199