In recent years, with the rapid development of robot technology, the application of robot has become more and more extensive. AGV, as a wheeled mobile robot, has the advantages of high reliability, strong adaptability and high degree of automation. It is widely used in the transportation and transmission of goods in modern manufacturing system and warehousing logistics system, and is one of the key equipment of modern warehousing system and flexible manufacturing system. Guidance and track tracking are the key technologies to realize AGV's autonomous movement. The method of laying predetermined track has the characteristics of convenient use and low cost, which is widely used in engineering practice. Due to the incomplete constraints between AGV wheels and the ground, it is difficult to establish a more accurate kinematic model. Many classic and mature control algorithms (such as PID control algorithm) are difficult to achieve. Fuzzy control is an algorithm based on practical experience and described by language. It does not need to establish an accurate mathematical model. It has strong fault-tolerant ability and is applied to non-linear systems Coupling system and other control systems have good results. According to the structural characteristics of forklift AGV, the corresponding fuzzy controller is designed to solve the magnetic navigation problem of forklift AGV in different paths.
1. Structure and motion analysis of forklift AGV
1.1 structure and driving part
The chassis structure of forklift AGV mainly includes five wheel type and three wheel type. AGV of three wheeled forklift truck can neither rotate nor translate, and its flexibility and stability are poor. Compared with the three-wheel forklift AGV, a universal wheel is added on the left and right sides of the steering wheel, which not only improves the performance of supporting balance, but also can assist steering, and the comprehensive performance is better than that of the two-wheel forklift. The relevant dimensions of the five wheel forklift AGV researched and developed here are shown in Figure 1
1.2 analysis on the movement and rectification of forklift AGV
Considering the safety of forklift AGV in operation, it is usually necessary to ensure that the forks are in front of people, while the driving wheels of AGV are located in the rear of the vehicle, which makes the movement mode of AGV different from other vehicles. Take the driving wheel and free wheel as the representative, analyze their motion characteristics, as shown in Figure 2. Firstly, it is assumed that there is no sliding friction between AGV and ground, and the ground is horizontal. When the AGV is rotating, for example, when the AGV needs to move up a certain displacement of the route. At this time, the driving wheel first rotates clockwise, then drives the AGV forward, drives the free wheel to the predetermined route, then the driving wheel rotates in reverse, and adjusts itself to the route. In this process, the motion track of AGV is adjusted to the ideal route by the swing effect of driving wheel. In the adjustment process, the distance along the ideal route is called adjustment distance adjustment distance, which is related to the speed of AGV free wheel, the rotation speed of driving wheel and the offset of two routes.
2. Navigation and control system of AGV
2.1 agv navigation mode
There are many ways of AGV navigation, and there are four common ways in China, ① Laser guidance uses laser ranging and angle to determine its own positioning, and realizes the movement according to the expected trajectory; ② visual guidance uses continuous acquisition of surrounding scene information to locate, while positioning in the process of movement, it establishes the scene map, and finally establishes the model of the whole environment, so as to realize the autonomous navigation; ③ inertial navigation uses the known position information, through the movement During the process, the integration of speed and orientation can obtain its own positioning, so as to achieve the purpose of navigation; ④ magnetic navigation can obtain the information of magnetic stripe by laying magnetic stripe on the preset path, and the magnetic sensor installed on the AGV can obtain the information of magnetic stripe, and through real-time correction, so as to ensure that the AGV can move in accordance with the preset track. Among the above methods, magnetic navigation is widely used in engineering practice because of its convenience and low cost, so the research is based on magnetic navigation.
2.2 hardware composition of control system
The control system is mainly composed of three parts, namely, perception part, decision-making part and execution part, as shown in Figure 3.
1) The sensing part is a STM32 development board, which is used to collect and transmit information about the deviation between the base and the magnetic stripe. There are mainly infrared sensors and ultrasonic sensors. After sensor information fusion, the environment around forklift can be identified.
2) The decision-making part takes STM32F103ZET6 embedded chip as the core chip of the control system. The chip is based on the core of Cortex-M3, which can effectively process the information from the sensing part and transmit the analysis and calculation results to the execution part, as shown in Figure 4.
3) The executive part is mainly composed of the steering wheel motor, the straight wheel motor and the hydraulic motor of the fork structure to realize the movement of the forklift and the delivery of goods.
2.3 software structure of control system
The AGV has two kinds of navigation control modes, which are self guidance control and manual operation control based on preset magnetic stripe. The architecture of the software control system is shown in Figure 5. Among them, manual control is relatively simple, driving AGV forward and backward through electronic control, which is used for AGV debugging and manual driving.
In the automatic control mode, on the one hand, it is necessary to obtain the deviation of AGV from the preset magnetic stripe in real time, and reduce the deviation in time through the correction program. On the other hand, based on safety considerations, when there is an accident or there is an obstacle on the magnetic stripe, the AGV should be able to automatically emergency brake, and recover the operation state after the obstacle is removed. In addition, in the actual project, straight-line driving is the main moving state of AGV. Considering the efficiency of the project, it is necessary to write acceleration and deceleration programs. Therefore, in the automatic control mode, it is necessary to write the automatic correction program, emergency braking program, acceleration and deceleration program, turning program and fork lifting program.
3. Fuzzy control algorithm design
Fuzzy control is a process of simulating people's thinking, that is, through observing and analyzing things, making judgments and decisions. L.A.Zadeh put forward the principle of incompatibility. As the complexity of the system increases, the clarity of the system will gradually decrease. When the threshold value is reached, the clarity and complexity will repel each other, resulting in a fuzzy control concave. Fuzzy control is a kind of nonlinear control, the core of which is to use the theoretical knowledge of fuzzy set to transform the control algorithm into a language that can be described by computer language. Because of the strong robustness of fuzzy control, it is not necessary to establish an accurate mathematical model, especially in dealing with nonlinear, strong coupling time-varying control systems.
The working principle of the fuzzy controller is to change the input clear signal into fuzzy quantity through the fuzzy interface, then through the fuzzy reasoning algorithm calculation, after the fuzzy reasoning machine processing, the fuzzy set is obtained, and finally the fuzzy set is changed into the clear quantity by the de fuzzy. The basic structure of fuzzy control system is shown in Figure 6.
The correction program is the core of the automatic control mode. The more reasonable the correction program is, the better the tracking performance of AGV will be. When the AGV runs on a straight line and a curve with a small turning angle, the correction program will be called.
The magnetic navigation sensor used in the AGV has 16 magnetic signal monitoring points, which are installed in the front of the AGV, 2 places away from the ground, and the selected magnetic stripe width is 3. In the vertical direction of the magnetic stripe, the magnetic sensor can connect four magnetic signals at the same time. In order to determine the position of AGV relative to the magnetic stripe, the monitoring points of the magnetic sensor are numbered. H0-h15 are 16 monitoring points respectively. The low potential 0 indicates that no magnetic signal is detected, and the high potential 1 indicates that the magnetic signal is detected. When the magnetic signal is detected, the indicator lights up, and all the potential information is stored in the register successively, occupying 16 bytes in total. In order to correct the deviation, we need to continue straight, left and right deviation. When the corresponding indicator lights of the four monitoring points in the middle are on, the corresponding potential is 1 at this time, indicating that AGV is in straight running state, as shown in Table 1.
Considering that the ground is not completely flat, a certain error range should be allowed. When the other indicator lights on the left and right sides of the four monitoring points in the middle are on, it is also considered as straight running state. According to the above analysis, the sensor potential can be fuzzed into three deviation States, as shown in Table 2.
In order to make the running track of AGV closer to the preset track, the output torque of the steering motor is different according to different offset degree when writing the program, and the steering in the correction program should be opposite to that in Table 2.
4. Path planning and track tracking test
4.1 test device and path planning
The AGV reconstructed in the laboratory is selected as the test object of track tracking, as shown in Figure 7.
In the figure, the magnetic sensor is located on the central axis of the AGV driving wheel, with a distance of 2cm from the ground. It is set with a guide path for straight travel, turning, forking, unloading and other positions, which can meet the use requirements of the general freight yard. According to the actual needs of the project, the magnetic guide wire is laid in advance, and the trajectory planning of the fork is shown in Figure 8. The empty AGV starts from stop a, moves along the magnetic stripe to position C, then AGV moves from position C to position D, and then goes straight to e to exit the warehouse, picks up the cargo at the bottom of the fork, lifts the cargo and returns to position F, then moves along the magnetic stripe to position F', and then returns to position a along the original route.
4.2 track tracking test
The track tracking of AGV is to let AGV move along the pre laid magnetic stripe. By adjusting the speed vector in real time, the AGV keeps moving above the magnetic stripe. According to the magnetic signal sensed by the magnetic sensor, the displacement offset can be obtained - 1-4 bits to the left and 1-4 bits to the right. According to the different offset, the steering motor outputs different steering torque and driving torque. In the performance test of AGV, Yao Jianyu's research method on the experimental AGV is used for reference. A 20cm × 3cm white marker strip (as shown in Figure 9) is pasted on the magnetic lead wire. In addition, the water-based brush is fixed at the center of the magnetic sensor, and the core of the brush just touches the marker strip to record the actual movement track of AGV.
After one-time Cross shipment, delivery and unloading, measure the notes on each sign strip, record the lateral offset distance of each sign strip's handwriting respectively, and the result obtained after matlab analysis is shown in Figure 10.
It can be seen from the figure that the deviation range of AGV is very small when it is just started, and the maximum deviation is 2cm; the deviation is still very small when it is on the straight road section; when it is on the turning road section, the deviation of AGV increases, and it starts to decrease and then increases when it reaches a certain range. Therefore, the forklift AGV is zigzag when it takes the turning road. It can be seen from the test results that the deviation range of AGV is ± 6cm, which can meet the deviation requirements of the design.
A fuzzy controller is designed for the motion and magnetic guidance characteristics of forklift AGV. Analyze and process the digital deviation signal obtained by magnetic sensor detection, output the result to the drive motor according to the fuzzy control rules; analyze the various road working tracks in the actual project, realize the autonomous navigation of AGV by programming, and improve the correction ability of AGV. The test results show that the forklift AGV based on fuzzy control has good track tracking performance.
 Yang Tianxu, Lou peihuang, Qian Xiaoming. Omnidirectional mobile AGV path tracking control technology with multi fuzzy controller [J]. Mechanical design and manufacturing engineering, 2016, 45 (7): 57-61
 Ren Yu, Zhao Shitao. Tracking control method of AGV deep reinforcement learning path [J]. Journal of Hangzhou University of Electronic Science and technology: Natural Science Edition, 2019, 39 (2): 28-34
 Chen D D，Shi Z Y，Yuan P J，et al. Trajectory tracking controlmethod and experiment of AGV[C]//2016 IEEE 14th InternationalWorkshop on Advanced Motion Control （AMC）.IEEE，2016：24-29.
 Ge Hongdou, he Lei, Cao Yaoqing, et al. AGV control based on Fuzzy PID control [J]. Automation of military industry, 2017, 36 (12): 76-79
 Kong fanwang. Design and application of modeling and trajectory tracking control methods based on different types of AGVs [D]. Hangzhou: Zhejiang University, 2018
 Wang Dianjun, Guan Xiangyu, Chen ya, et al. Kinematic analysis and Simulation of two-way dual drive AGV robot [J]. Manufacturing automation, 2016, 38 (3): 42-46, 56
 Zhou Xiaojie. Development trend of AGV system and guidance mode [J]. Nanfang agricultural machinery, 2017, 48 (4): 80
 Qin Shanghuo. Research and development of magnetic navigation forklift AGV control and guidance system [D]. Nanning: Guangxi University, 2017
 Gu Jun, Zhang Yu. Application status and development trend of fuzzy control [J]. Chemical automation and instrumentation, 2017, 44 (9): 811-812902
 Xiong B，Qu S R. Intelligent vehicle’s path tracking based onfuzzy control[J].Journal of Transportation Systems Engineering andInformation Technology，2010，10（2）：70-75