Based on the analysis of the domestic and foreign market and technology status of Automated Guide Vehicle (AGV), a low cost AGV suitable for factory workshop is designed.An AGV fuzzy controller based on attitude research is designed by using ATmega128 single chip computer. Fuzzy control algorithm is used to control the two-wheel differential steering of AGV. The optimization of the driving state of the car based on attitude analysis is proposed, and the motion tests on different roads are carried out.Experiments show that the fuzzy controller with attitude analysis can pre-judge the attitude of some special running points of the car, ensure the attitude of the car more effectively during the AGV driving process, reduce the overshoot caused by the car running, and make the car track the guidance path more accurately.In recent years, with the increasing application of flexible manufacturing and assembly systems in China’s industrial field, AGV has been paid more and more attention by all walks of life.However, the development of AGV in China started late, and technology generally lagged behind the European and American countries. The products were slower to update, the core competitiveness was relatively small, and a large number of key technologies were controlled by European and American enterprises.
In China, the development of low-end AGV products was more competitive.This kind of product does not require powerful functions, but only meets the normal handling requirements.Therefore, how to effectively control the transport and operation of AGV has become the core issue of the development of this kind of products [3].At present, there are many colleges and universities studying AGV technology in our country, but most of them are advanced AGV technology research.
Only a few of them involve this kind of low-cost AGV research.Aiming at the low cost AGV which is more suitable for the Chinese market, the research and design of the product is carried out based on the actual project of an enterprise.The AGV fuzzy controller is designed and optimized by combining the fuzzy control theory with the special attitude analysis method.The AGV prototype is guided by magnetic strip and operates by differential steering.The experimental data and results are obtained by the prototype.The control system of AGV consists of three parts: path identification system, speed control system and main control system.Path recognition system is the “eye” of AGV, which can accurately tell the position of the car.The speed control system includes two parts: the speed detection system and the motor drive system.The main control system is the core of intelligent vehicle, which is composed of microcontroller and peripheral circuit.In this design, the magnetic navigation sensor is used to read magnetic strips to reflect the driving position and migration status of the car. The magnetic strip navigation sensor has eight detection points and eight digital signal output ports.The main control unit obtains the motor control quantity of the car’s offset and speed control system by analyzing the information obtained by the magnetic navigation sensor, so as to achieve the purpose of accurate control of the intelligent vehicle.The speed control system of the prototype adopts the brushless DC motor and the D-type driver, which control the speed of the left and right wheels of the prototype respectively.Fig. 1 is the overall control plan of the smart car.The design of fuzzy controller includes the fuzzification of input and output parameters, the establishment of fuzzy rule base, fuzzy reasoning and the de-fuzzification of output parameters.The fuzzy controller designed in this project sets input parameters as deviation e of magnetic navigation sensor and deviation change rate e, and output parameters as duty cycle difference U of PWM wave of left and right motor speed control signals.The deviation rate e is defined as the domain absolute value of the last deviation minus the domain absolute value of this deviation, that is, e = | Elast_time |-| Ethis_time |.The fuzzy controller is used to analyze and encode the signals obtained by sensors, so as to form fuzzy quantities and participate in fuzzy reasoning.The 8-bit digital signal from the magnetic strip navigation sensor, whose 8-bit output signal is L7 to L0 from left to right, indicates the deviation of the car relative to the magnetic strip.Because the deviation signal is discontinuous and does not show a monotonic increasing law, the coding value is set.Its coding table is shown in Table 1.Set the ambiguity of the deviation value as E, the ambiguity of the deviation change rate as EC, and U as the speed change control.After analysis, the basic domain of deviation e is [-4,4], the basic domain of deviation change rate e is [-1,0], and the basic domain of control quantity u is [-4,4].{Negative large, negative medium, negative small, 0, positive small, positive medium, positive large} is {NB, NM, NS, ZO, PS, PM, PB}.
Above: u indicates the difference of duty cycle of PWM wave between left and right speed regulation.
It is stipulated that when U > 0, the duty cycle of PWM wave signal of left motor is large, indicating that intelligent vehicle should turn left; when U < 0, the duty cycle of PWM wave signal of right motor is large, indicating that intelligent vehicle should turn right.Because the fuzzy controller needs the fuzzy quantity, and the sensor detects the precise quantity, it needs to fuzzify the precise quantity.The deviation E and output u are fuzzified by triangular membership function, as shown in Figure 2 [5].For the fuzzification of deviation change rate e, single-valued fuzzification method is adopted, that is, EC is -1, corresponding to increase (PB); EC is 0, corresponding to no change (ZO); EC is 1, corresponding to decrease (NB) [6].If E and EC then U.Among them, E and EC are the fuzzy sets of fuzzification of input system deviation E and deviation change rate e, respectively.Through the analysis of the route, the corresponding fuzzy control rules are obtained according to the driving experience as shown in Table 2.The establishment of AGV fuzzy control rule base should satisfy the following requirements: when the deviation is large, the speed difference between left and right motors should be large, so that the car can quickly reduce the deviation; when the deviation is small, while making the car reduce the deviation, it should try to avoid excessive overshoot, so as to avoid introducing reverse deviation [7].The center of gravity method is used to solve the ambiguity. For discrete universe, the method can be expressed as formula (1) [8].After establishing the model of the fuzzy controller, the fuzzy controller is simulated and analyzed by using the Fuzzy toolbox in matlab.Figure 4 shows.According to the response curve, the system has fast response and strong anti-interference ability.Attitude analysis is an analysis of the angular deviation between the center line of the car body direction and the magnetic strip during AGV operation, as shown in Fig. 5.When running in a straight line, the deviation angle is the angle between the center line of the car body and the magnetic strip; when running in a curved road, the deviation angle is the angle between the center line of the car body and the tangent line of the magnetic strip.For quantitative analysis of precise angle deviation, at least two magnetic bar navigation sensors are needed.The accurate angle deviation of the car relative to the magnetic strip can be calculated by the deviation values read by the sensors placed at the front and rear ends of the car body.In this design, there is only one magnetic strip sensor, which can not achieve accurate quantitative analysis. The analysis of angle deviation is simplified by AGV kinematics model, so as to achieve the initial attitude correction.The AGV kinematics model is shown in Fig. 6.In the figure, VL and VR are the speed of left and right driving wheels, m/min; L is the distance between driving wheels, m; R is the rotation radius of AGV, M.The kinematics model and formula (3) show that the actual running process of the car is the combination of straight line and arc.At the same time, the deviation correction of the car is actually to interpolate the linear track and the arc track by using the arc segment.When straightway runs, the deviation correction uses arc to interpolate the straight line, so the attitude will change while correcting the deviation.In the design of fuzzy control, the straight-way operation should be completed in 6 test points (2 out of 8 test points).Therefore, the deviation is NS or PS, and the deviation change rate is NB, which is the attitude correction point, as shown in the light gray area of Table 2.When the car runs from NS position to NM position or from PS position to PM position, the car runs along the direction of deviation from the magnetic track; when the arc represented by formula (3) is used for attitude correction, the correction trajectory is an arc, in the course of correction, the distance from the magnetic track of the car will increase initially, and then gradually decrease.Taking the deviation from NS (slightly right deviation) to NM (middle right deviation) and the correction from NM position to NS position as examples, the sketch diagram of the car straightway deviation correction process is shown in Fig. 7.In Figure 7, the critical point of NS and NM to the critical point of NM and NS is the car deviation correction stage. At the end of the correction, the car and the magnetic strip have a large angle difference and form a bad attitude.Formula (2) shows that if the speed of the left and right driving wheels meets the requirements of Formula (4) ~ (6) when the car is turning, thermostatic element the running speed of the car in straight line and circular arc motion is constant v.In order to improve this bad attitude, another arc opposite to the deviation correction arc can be used for attitude correction.The attitude correction process not only improves the car’s attitude, but also makes the car approach the magnetic strip further.The attitude correction process is shown in Figure 8.
Formula (8) shows that the car can form a reverse attitude correction arc with a given suitable D.The value of D is related to both L and Dbase. In the actual design process, the distance between the left and right driving wheels of the car is fixed, the appropriate D is selected by Dbase, and the correction time is set to 1 s.D should be set appropriate value so that the end point of car attitude correction can be located in ZO or NS (PS) area to avoid unnecessary overshoot [10].In this design, in order to observe the effect of attitude correction, the larger value of L is fixed at 80 cm.When Dbase 30/25511.76%, v7.39 m/min, Dz 13/2555.10%.When Dbase 40/25515.69%, v9.85 m/min, Dz 17/2556.67%.When Dbase 50/25519.61%, v12.31 m/min, Dz 21/2558.24%.The attitude optimization method of curve operation is the same as that of straight operation, but the arc section is interpolated by the arc section in the process of curve operation.Bend operation is completed by two outermost points on the left and right sides, totaling four points.Therefore, the deviation is NM or PM, and the deviation change rate is NB. These two points are bend attitude correction points, as shown in Table 2 in dark gray area.Taking NB point as an example, the bend attitude correction is shown in Figure 9. The bend analysis process and parameter setting are not discussed.In this design, the “fixed value” method is used to optimize the attitude, that is, the radius of the reverse correction arc and the correction time are fixed, so the deviation in the correction process may show a trend of oscillation attenuation [12].In the experiment, it is found that in the course of attitude and deviation correction, the number of oscillations is generally 1-3 times, and then the attitude of the car returns to an ideal state.Taking the straight line operation as an example, the cause of this phenomenon is shown in Figure 10.The deviation correction trajectory is tangent to the car body center line. Figure 10 shows that the attitude angles of the car corresponding to the three trajectories are different at the critical points of NS and NM, which leads to the different attitude angles at the critical points of NM and NS.If the same radius of the correction arc and the same correction time are used to correct the attitude of the car with different initial attitude angles, the oscillation of deviation will inevitably occur.
This oscillation phenomenon indicates that when attitude correction is carried out, redundant and insufficient corrections are given, so it may need 1 to 3 normal corrections to make the car achieve satisfactory posture.For this design of low cost solution, this phenomenon is more difficult to avoid, but has no effect on normal use.The driving system structure of the prototype is shown in Figure 11.The motor and the driving system are connected by the synchronous belt of 5M600 type, which can realize the positive and negative rotation. At the same time, its vibration reduction function can help to reduce the impact force when the car is parked in emergency.The driving wheel is cast iron core rubber wheel of 26.7 cm, and the motor is a brushless DC motor of FBL-92H25301RS of a company. The rated torque of the motor is 0.8 N.m, and the rated speed is 3 000 r/min.The helical gear reducer with a deceleration ratio of 1:30 is selected as the reducer, and its model is 6GU-30K.AGV frame is welded with No. 10 and No. 5 channel steel, as shown in Fig.
12.In the experiment, the N-pole magnetic strip with a width of 30 mm and a thickness of 1.2 mm was selected. The distance between the magnetic strip navigation sensor and the magnetic strip was 20 5 mm.At the same time, the test ground is terrazzo ground, and the rolling friction coefficient with rubber wheel is 0.015-0.020, which meets the test requirements.The load-carrying capacity of the test car is about 60 kg.In the experiment, the stopwatch is used to time the running time of the car on the 90 degree bend of the specified length or radius. The actual running speed of the car can be calculated by counting the time. The actual running speed is compared with the set speed, that is to say, the running attitude of the car can be known.The reason is that if the actual running speed of the car is much less than the set speed, the attitude of the car is not good, and the “snake” runs too much, which leads to the increase of the actual running distance of the car and the decrease of the average running speed.The test is divided into straight test and bend test.In order to better observe the effect of car attitude correction, the distance L between driving wheels is fixed to a larger value, and the setting speed of the car is reduced, so that the turning radius R of the car increases, and the attitude correction process is more specific and intuitive.Straight track test intercepts 8 m long magnetic strip as guidance path, and divides it into two groups at different setting speeds and carries out six experiments.After calculating the average speed of each group, the average speed of each group is compared with the set speed, and the relative error limits corresponding to the average speed of each group are calculated.The straight-way test scheme and data are shown in Table 3.According to the data in Table 3, the relative error limits of the average speed measured by each group of straight track running of the prototype car are within 10%, and the relative error limits can be guaranteed within 5% when the speed is set low.At the same time, under the same set speed, the measured actual speed is not much different, and the running attitude of the car is guaranteed to be better.The test method of bend test is the same as that of straight test.Bend test takes 90 degree bend with specified radius as guidance path. The test scheme and average data are shown in Table 4.The driving time, actual speed and relative error limits in Table 4 are averages.In bend test, the relative error limits of each group are all within 15%, and when the speed is low, the relative error limits can be guaranteed within 8%.At the same time, it is observed that the overshoot is the largest when the car enters the bend during its operation, and the overshoot decreases gradually as the car continues to move, so the attitude of the car in the bend operation is also guaranteed to be better.The minimum radius of bend can reach 1.5 m in the test.A fast and simple optimization method for the low cost AGV controller is proposed. The traditional AGV fuzzy controller is designed and optimized by analyzing the AGV running attitude and combining with the fuzzy control theory.In this paper, the AGV prototype is designed, and the actual test of the controller is carried out.In order to observe the attitude correction effect of the car, the prototype and the test scheme are processed.
The treatment scheme includes fixing the designed driving wheel spacing L to a larger value, reducing the setting speed of the car, and setting the test distance and the angle of the test bend to a fixed value.
Experiments show that the fuzzy controller based on attitude analysis can effectively guarantee the running attitude of AGV in straight and curved lanes, improve the running accuracy of AGV and reduce the overshoot caused by running.