Study Of AGV Agent Control System

A practical AGV(Automated Guided Vehicles) agent control system based JADE(a wide used Java agent platform, Java Agent Development Environment) is present in this paper, to fit in with requirements of intelligent AGV technology and development of agent standards. Some middlewares were developed to extend the functions of Java Agent, and help Agent to interact with motion control card, OPC server, and database. Modules about real-time trajectory generation, motion control, pose estimation and safety control are designed in PMAC motion control card, and management modules about path planning and database in PC 104 industrial control computer. Human-machine interface Agent can start in other computers. This control system is decentralized control system, can improve openness, flexibility and intelligence of AGV and AGV system. It is easy to communicate AS/RS(Automated Storage and Retrieval System) control systems.Two methods of trajectory planning between two route points are developed because common method of global path planning is difficult to adapt to the requirements of AGV agent control system. Formulas of quintic Hermite polynomial curve and parametric curve are derived to ensure the continuous curvature of the path, so that AGV steering angle can change continuously while tracking, and AGV can reach the target point in the right position and orientation. The speed and steering control method is proposed. Although the two methods have their own advantages and disadvantages respectively, both of them can be used to generate trajectory online and easy to guidance control. The results of simulation and experiments are given.AGV inertial navigation system is designed which composes of several sensors, such as encoders, gyroscope, magnetic ruler, and a simplified Kalman filter, an extended Kalman filter. Simplified Kalman filter is used to detect orientation changes by fusing data from encoders and gyroscope. The algorithm can be easy programmed in motion control card, and measurement accuracy of attitude angle improved. Pairs of magnets are installed in the path at certain interval and are detected by magnetic ruler to correct accumulative error of position and orientation. AGV's position can be measured after linear scanning a pair of magnets and orientation can be calculated with the distance between AGV's center and magnetic nails. An EKF(extended Kalman filter) is used to get the more precise position and orientation by fusing data from Dead-Rankoning system and magnets correcting system. In order to improve computational efficiency, matrix computing of EKF is programmed in PC 104 industrial control computer. The experiments prove that the integrated navigation system can meet the requirements. The navigation system has advantages of high positioning accuracy, low cost and convenient to install.It is difficult to get the absolute position of magnets because of obstacle in job-shop and this often causes trouble in debugging AGV. In order to rapidly layout AGV, teach-in method is used to measure and record the location of magnets, and iterative learning control is introduced to update the aim point of AGV. After several times learning, AGV can run over magnets accurately. Accumulation of error can be eliminated when correction is done. The results of experiments are given to demonstrate the efficiency. Q learning method is used in intelligence planning path with magnets to achieve the shortest path search, obstacle avoidance, task scheduling and so on. Q learning algorithm is convenient and practical, need not establish a precise model of the environment, and can simplify agent program.It can construct a new kind of practical AGV control system with the methods of AGV Agent of control system, trajectory planning, guidance control, navigation system and intelligent path planning with magnets in the paper.