Study Of The Intelligent Control System For Industrial Application

With the rapid progress in Internet of Things(IOT)technology and robot technology,the industrial equipment and agricultural equipment have been greatly improved in terms of intelligence,accuracy and efficiency.It is also pointed out in the national ?"thirteenth five-year plan" for agriculture and Industry 4.0 that we should effectively improve industrial and agricultural technologies and equipment and increase the information technology level in order to create a strong intelligent industry pattern.At present,the industrial AVG and agricultural plant protection multi-agent systems are the most representative applications;therefore,in-depth studies on the intelligent control systems for industrial application are of epoch-making significance to building an intelligence industry pattern in the Internet era.In this paper,we combine theories with practices,and study the intelligent agent control system in a bottom-up approach from the bottom controller,the intermediate network layer to the upper decision-making layer according to the actual requirements for industrial and agricultural multi-agent systems.The main contents are as follows: In Section 1 – Introduction,we start by analyzing the current application researches and development trend,list the problems with multi-agent systems in the current intelligence industry applications,and describe in detail the developments in the theoretical methods and applications of intelligent agent control in two scenarios-factory delivery system and agricultural plant protection.Then we analyze and discuss several key problems based on the recent progress in relevant researches at home and abroad,and introduce the significance and main contents of this paper.In Section 2 – Supporting Research for the Application of Intelligent Agent Systems,we mainly describe the test conditions for actual operation and supporting research in combination with our horizontal scientific research projects.We propose a ?mixed multi-agent management system structure?,and introduce the basic design concepts and main components of the system from the perspective of system framework and implementation,and then we build hardware and software systems on this structure.We design three intelligent agent system models,which can be changed to heterogeneous multi-agent systems to fit specific scenarios,serving as an example to show how to turn the theories in this paper into practice.Unlike the experimental systems generally for theoretical studies,we build embedded intelligent agent controllers into these systems and conduct a lot of data processing and real-time controller studies in the resource-restricted environment.In Section 3 – Study on the Stability of the Agent controller based on the 2D Markov Jump System,by studying the design problems of the robust controller of an intelligent agent system under the 2D Markov jump system,we try to eliminate the impacts of the non-ideal agent model on the agent itself,such as the uncontrollable situations caused by manufacturing deficiencies and uncertain environments in actual practice.We use the local state-space model of Fornasini-Marchesini to describe its uncertainty,ensure that the transition mode probability matrix is uniform and known,and propose a pattern-dependent state feedback controller design.Compared with similar studies,we consider the impacts of random variables,external disturbances and uncertainties,and obtain the controller gain by solving a set of linear matrix inequalities,which has optimized the constraints and reduced the computational complexity and hardware dependency.In Section 4 – Study on the Predictive Control of the Intelligent Agent System in a Network with Time-varying Delays and Packet Losses,we mainly study the design of a robust output feedback prediction controller for a kind of constrained linear system affected by periodic measurement packet losses and external disturbances.Unlike the typical Luenberger observer,the observer we design first ensures that the system is asymptotically stable and then it converges the continuous packet losses and other errors into a bounded compact set,which decouples the dynamic equations of state estimation system and the error system and thus improves the performance of the controller and ensures the iteration feasibility of the algorithm.By considering the extent of the impact of periodic measurement losses,we determine the exact convergence of the system state.In Section 5 – Study on the Target Location of the Intelligent Agent System in Case of Random Network-induced Delay,according to actual application needs,we design two kinds of optimized agent spatial location and tracking algorithms from the agent's first person view and the upper controller's third-person view,which simplify the constraints or perturbation conditions of the control algorithm and improve its efficiency and accuracy.The integration of sensor data can not only effectively improve the measurement accuracy,but also solve the coverage limitations of a single sensor.In combination with the study mentioned in Section 4,we install a Markov link delay suppressor on each end of the controller communication module to suppress random delays.From this,we propose an improved asynchronous data fusion method,by which the controller can still have high control performance in case of a large time delay.This method can provide reference for the selection of agents and parameter design in actual practice.In Section 6 – Study on the Cooperative Control Methods for Heterogeneous Agents under Uncertain Constraints,we mainly state that the agents and their communication control system is the important foundation for multi-agent coordination controls and applications,and discuss a ?lead-follow? multi-agent cooperative control method based on sliding mode control.We fit the system uncertainties with the optimized multi-input and multi-output compensator and realize the robust formation control of multiple agents by using both the fuzzy compensator and the sliding mode controller.For the problem with swarm intelligence control,we propose a trend control strategy based on image morphology as the countermeasure.In order to deal with the computational cost and maintain the real-time performance of the above algorithm,we discuss and design a special embedded parallel processing system.This method is proposed for the first time.The multi-agent system applied in industry and agriculture involves multiple layers and a number of key issues,which should not be studied separately.Therefore,in this paper,we propose and design for the first time a multi-agent mixed control system framework and then study the issues on each layer of the framework.Main contributions of this paper are as follows:(1)this paper proposes a robust controller for the intelligent agent system under the 2D Markov jump system to effectively overcome the impacts of the agent design and operating environment and ensure the usability and stability of individual agents;(2)it proposes a robust output feedback predictive controller to eliminate the impacts of time delay and packet losses in the multi-agent communication and ensure the effective connection between the agents on the bottom layer and the upper decision-making layer;(3)then it proposes anoptimized asynchronous sensor integration method to improve the measurement accuracy of the sensor and further optimize the performance of controllers on each layer;(4)At last,it proposes a cooperative control method for the application layer based on the fundamental multi-agent research and its experimental platform and makes an attempt in swarm control,which can be used as an important reference.The research results of this paper provide a new idea for the research on the application of agent cooperation in industry and agriculture and can be used as an important reference,especially for the application of multi-agent system control theories in practice.