Logistics classification and sorting control method and system based on fuzzy control

Abstract

The present invention provides a logistics classification and sorting control method and system based on fuzzy control, which relates to the field of intelligent control technology, including obtaining multidimensional attribute information of items to be sorted, and obtaining a standardized item attribute feature vector; based on the item attribute feature vector, constructing a fuzzy set describing different sorting categories, constructing a composite membership function for the fuzzy set of each item attribute dimension, inputting the multidimensional attribute information of the item into the composite membership function of the corresponding attribute dimension, and determining the best sorting category of the item; constructing a complete fuzzy control rule base, adjusting the position, speed and acceleration of the actuator in real time, and optimizing the item sorting trajectory.

Description

Technical Field

[0001] The present invention relates to intelligent control technology, and in particular to a logistics classification and sorting control method and system based on fuzzy control. Background Art

[0002] Traditional logistics sorting mainly relies on manual work, which has problems such as low efficiency, high error rate, and high labor intensity. In order to improve sorting efficiency and accuracy and reduce labor intensity, it is necessary to research and develop intelligent and automated logistics classification and sorting control technology.

[0003] At present, logistics classification and sorting control mainly uses barcode recognition, RFID recognition, machine vision and other methods to obtain item information, and controls the sorting equipment to send items to designated areas or containers through pre-set sorting rules and algorithms. Existing sorting control methods usually use pre-set sorting rules and simple PID control strategies, which are difficult to adapt to complex sorting environments with a wide variety of items and different sizes and weights. The sorting efficiency and accuracy need to be further improved. At the same time, the operating parameters of the sorting equipment are difficult to optimize in real time, the trajectory planning is not flexible enough, and the control of sorting time and energy consumption is not fine enough. Summary of the invention

[0004] The embodiments of the present invention provide a logistics classification and sorting control method and system based on fuzzy control, which can solve the problems in the prior art.

[0005] According to a first aspect of the embodiments of the present invention,

[0006] Provides a logistics classification and sorting control method based on fuzzy control, including:

[0007] Acquire multi-dimensional attribute information of the items to be sorted, collect attribute data of the items through a multi-sensor array set at different positions of the sorting line, and pre-process the collected attribute data to obtain a standardized attribute feature vector of the items;

[0008] Based on the feature vector of item attributes, a fuzzy set describing different sorting categories is constructed. For the fuzzy set of each item attribute dimension, a composite membership function is constructed, and the continuity of the composite membership function at the boundary of adjacent sub-intervals is achieved through smooth transition technology; the statistical characteristics of the distribution density and discrete degree of the item attribute data points in each sub-interval are comprehensively considered, and the basic function type and corresponding parameters of the composite membership function in each sub-interval are adaptively determined to establish a fuzzy mapping relationship between item attributes and sorting categories; the multi-dimensional attribute information of the item is input into the composite membership function of the corresponding attribute dimension, the membership value of the item on each fuzzy set is calculated, and a fuzzy classification rule base is constructed. The fuzzy reasoning method is used to comprehensively evaluate the membership value of the item to obtain the membership degree of the item to each candidate classification; according to the membership degree of the item to each candidate classification, the final classification result of the item is determined through defuzzification processing, and the best sorting category of the item is determined;

[0009] According to the optimal sorting category of items, a complete fuzzy control rule base is constructed, and the Mamdani reasoning and center of gravity defuzzification methods are used to implement the software algorithm of the fuzzy controller; actuators are selected to realize the sorting action of items, and according to the control command signal output by the fuzzy controller and the physical parameter model of the actuator, fuzzy PID control and fuzzy sliding mode control algorithms are used to adjust the position, speed and acceleration of the actuator in real time to optimize the item sorting trajectory.

[0010] In an optional embodiment,

[0011] Based on the item attribute feature vector, a fuzzy set describing different sorting categories is constructed. For the fuzzy set of each item attribute dimension, a composite membership function is constructed. The steps of achieving the continuity of the composite membership function at the boundary of adjacent sub-intervals through smooth transition technology include:

[0012] Obtain multidimensional attribute data of the items to be sorted, and divide the attribute dimension into several non-intersecting sub-intervals according to the value range and data distribution pattern of each attribute dimension, each sub-interval corresponding to a fuzzy set; wherein the division of sub-intervals comprehensively considers the distribution density and discrete degree characteristics of the attribute data in this dimension;

[0013] For each fuzzy set of each attribute dimension, one or more function types are selected from a preset candidate basic membership function library as the basic membership function of the subinterval according to the clustering characteristics of the data points in the corresponding subinterval; the candidate basic membership function library includes a triangular function, a trapezoidal function and a Gaussian function;

[0014] By using the least square method and gradient descent parameter estimation method, the attribute data samples in the subinterval are used as the training set, and the optimal parameter set of the selected basic membership function is adaptively estimated by optimizing the objective function, so as to minimize the fitting error of the data distribution in the subinterval.

[0015] The basic membership functions of all sub-intervals of each attribute dimension are combined and integrated to construct a composite membership function on the attribute dimension. The basic membership functions of adjacent sub-intervals are locally adjusted and optimized by using the smooth transition technology of interpolation fitting and spline function approximation.

[0016] The multi-dimensional attribute feature vector of the item is input into the composite membership function of the corresponding attribute dimension, the membership value of the item on each fuzzy set is calculated, and the nonlinear mapping relationship between the item attribute and the fuzzy classification is constructed.

[0017] In an optional embodiment,

[0018] The steps of using the least squares method and gradient descent parameter estimation method, taking the attribute data samples in the subinterval as the training set, and adaptively estimating the optimal parameter set of the selected basic membership function by optimizing the objective function so as to minimize the fitting error of the data distribution in the subinterval include:

[0019] For a certain attribute dimension of the item to be classified, all attribute data points in each subinterval after the attribute dimension is divided are obtained, and the attribute data points are used as a training sample set, where each sample point includes the attribute value of the point and a preset true membership value;

[0020] According to the type of the selected basic membership function, an optimization objective function is constructed with the goal of minimizing the sum of squares of fitting errors. The objective function is used to measure the degree of fit of the basic membership function to the sample points in the subinterval.

[0021] Initialize the parameter vector of the basic membership function. For the triangular function and trapezoidal function, the parameter vector includes the left and right boundaries and the peak point. For the Gaussian function, the parameter vector includes the mean and standard deviation.

[0022] The constructed objective function is iteratively optimized using the gradient descent method. The gradient vector of the objective function with respect to each parameter is calculated through partial derivatives, and the parameters are updated along the negative gradient direction until the objective function value converges or the maximum number of iterations is reached.

[0023] The parameter vector at convergence after iterative optimization is used as the optimal parameter estimate of the basic membership function in this subinterval, so as to minimize the fitting error of the attribute data distribution in this subinterval and obtain the basic membership function after parameter optimization.

[0024] In an optional embodiment,

[0025] The steps of combining and integrating the basic membership functions of all subintervals of each attribute dimension to construct a composite membership function on the attribute dimension, and using the smooth transition technology of interpolation fitting and spline function approximation to locally adjust and optimize the basic membership functions of adjacent subintervals include:

[0026] According to the order of sub-intervals, the basic membership functions after parameter optimization on each sub-interval are sequentially spliced ​​to form a piecewise function covering the entire attribute value range;

[0027] For the intersection points of adjacent sub-intervals, a smooth transition technology of interpolation fitting or spline function approximation is used to construct a transition function that meets the continuity and smoothness conditions;

[0028] The smoothed piecewise function and transition function are integrated to construct a composite membership function covering the entire attribute dimension, so that it is consistent with the optimized basic membership function within the sub-interval and achieves continuous and smooth transition at the junction of adjacent sub-intervals.

[0029] On the premise of maintaining continuity and smoothness, the local shape parameters of the composite membership function are fine-tuned, an additional correction function is introduced, appropriate regularization constraints are set, and numerical optimization methods are used to locally adjust and optimize the composite membership function.

[0030] In an optional embodiment,

[0031] The steps of inputting the multi-dimensional attribute information of the item into the composite membership function of the corresponding attribute dimension, calculating the membership value of the item on each fuzzy set, building a fuzzy classification rule base, and using the fuzzy reasoning method to comprehensively evaluate the membership value of the item to each candidate classification include:

[0032] The multi-dimensional attribute information of the items to be sorted is extracted and input into the composite membership function of the corresponding attribute dimension. The membership value of the fuzzy set corresponding to each attribute dimension of the item is calculated by the composite membership function. The membership value indicates the degree of belonging of the item to the fuzzy set corresponding to the attribute dimension.

[0033] Construct a fuzzy classification rule library. The fuzzy classification rule is composed of the combination conditions of the attribute membership values ​​of the items as the antecedent and the classification results of the items as the consequent, which is used to describe the fuzzy mapping relationship between the item attributes and the classification categories.

[0034] For each fuzzy classification rule, the attribute membership value of the item is substituted into the antecedent part of the fuzzy classification rule, and the trigger strength of each fuzzy classification rule is calculated by the fuzzy reasoning method. The trigger strength is calculated by the combination of the attribute membership values ​​of each item, which indicates the degree to which the fuzzy classification rule is satisfied by the item attributes.

[0035] For each candidate classification category, summarize the reasoning results of all fuzzy classification rules pointing to this category, and take the maximum trigger strength of the item in all fuzzy classification rules pointing to this category as the item's membership degree to this category. The membership degree represents the belonging measure of the item to this classification category based on fuzzy reasoning.

[0036] According to the degree of belonging of the items to each candidate classification category, a decision strategy is adopted to determine the final classification of the items, and the items are divided into the classification category with the highest degree of belonging.

[0037] In an optional embodiment,

[0038] According to the best sorting category of items, a complete set of fuzzy control rule base is constructed, and the steps of implementing the software algorithm of the fuzzy controller using Mamdani reasoning and center of gravity defuzzification method include:

[0039] Input the input variable value of the current control cycle into the fuzzy controller to trigger the corresponding fuzzy control rule. According to the input variable language value combination of the antecedent part of each fuzzy control rule, calculate the membership value of each input variable corresponding to the language value, and use the minimum operation to obtain the triggering strength of each fuzzy control rule.

[0040] According to the triggering strength of each fuzzy control rule, the membership function of the output variable of the consequent part of the fuzzy control rule is truncated to obtain the membership function of the output variable under each fuzzy control rule;

[0041] The membership functions of all output variables that trigger the fuzzy control rules are combined, and the total membership function of the output variables is obtained by taking the largest operation;

[0042] The centroid defuzzification method is used to convert the total membership function of the output variable into an accurate control quantity. The centroid defuzzification obtains the defuzzified value by calculating the center of the area under the total membership function. In the actual calculation, the total membership function is discretized into a finite number of points by numerical integration, and then the centroid is approximately calculated by weighted average.

[0043] The numerical value obtained by defuzzifying the center of gravity is converted into a specific actuator control instruction, and the control instruction is sent to the actuator.

[0044] In an optional embodiment,

[0045] The actuator is selected to realize the sorting action of the items. According to the control command signal output by the fuzzy controller and the physical parameter model of the actuator, the fuzzy PID control and fuzzy sliding mode control algorithms are used to adjust the position, speed and acceleration of the actuator in real time. The steps of optimizing the item sorting trajectory include:

[0046] Establishing a kinematic model and a dynamic model of the actuator, establishing the forward and inverse kinematic equations of the actuator by using the DH parameter method, describing the mapping relationship between the joint angle and the end position and posture, and considering the dynamic characteristics of the actuator to establish the dynamic equation of the actuator;

[0047] Determine the optimization objective function and constraint conditions of the sorting trajectory, wherein the optimization objective function includes minimizing the sorting time, maximizing the average speed and minimizing the acceleration change, and the constraint conditions include the upper and lower limits of the joint angle, angular velocity and angular acceleration of the actuator and the boundary constraint of the end position; discretize the sorting trajectory, generate smooth trajectory segments between adjacent discrete nodes by interpolation method, and transform the continuous trajectory planning problem into a discrete optimization problem;

[0048] A fuzzy PID control algorithm is applied, with position error and speed error as input, and the parameters of the PID controller are adjusted online through fuzzy reasoning and defuzzification calculation to generate optimized position and speed control instructions, and the optimized position and speed instructions are substituted into the kinematic model of the actuator to calculate the angle and angular velocity of each joint; a fuzzy sliding mode control algorithm is applied, with acceleration error and acceleration change rate as input, and the sliding mode surface and control law are adjusted online through fuzzy reasoning and defuzzification calculation to generate optimized acceleration control instructions, and the optimized acceleration instructions are substituted into the dynamic model of the actuator to calculate the driving torque of each joint;

[0049] The sorting trajectory is iteratively optimized, and the position, velocity and acceleration instructions obtained by the fuzzy PID control algorithm and the fuzzy sliding mode control algorithm are used as the optimal values ​​of the current discrete node, substituted into the next discrete node, and the fuzzy PID control and fuzzy sliding mode control are repeatedly executed until the optimization of the entire sorting trajectory is completed.

[0050] According to a second aspect of the embodiments of the present invention,

[0051] Provide logistics classification and sorting control system based on fuzzy control, including:

[0052] The first unit is used to obtain multi-dimensional attribute information of the items to be sorted, collect attribute data of the items through a multi-sensor array set at different positions of the sorting line, and pre-process the collected attribute data to obtain a normalized attribute feature vector of the items;

[0053] The second unit is used to construct fuzzy sets describing different sorting categories based on the feature vector of item attributes, construct a composite membership function for the fuzzy set of each item attribute dimension, and realize the continuity of the composite membership function at the boundary of adjacent sub-intervals through smooth transition technology; comprehensively consider the statistical characteristics of the distribution density and discrete degree of the item attribute data points in each sub-interval, adaptively determine the basic function type and corresponding parameters of the composite membership function in each sub-interval, and establish a fuzzy mapping relationship between item attributes and sorting categories; input the multi-dimensional attribute information of the item into the composite membership function of the corresponding attribute dimension, calculate the membership value of the item on each fuzzy set, construct a fuzzy classification rule base, and use fuzzy reasoning method to comprehensively evaluate the membership value of the item to each candidate classification; according to the membership degree of the item to each candidate classification, determine the final classification result of the item through defuzzification processing, and determine the best sorting category of the item;

[0054] The third unit is used to build a complete fuzzy control rule base according to the optimal sorting category of items, and use Mamdani reasoning and center of gravity defuzzification method to implement the software algorithm of the fuzzy controller; select actuators to realize the sorting action of items, and use fuzzy PID control and fuzzy sliding mode control algorithms according to the control command signal output by the fuzzy controller and the physical parameter model of the actuator to adjust the position, speed and acceleration of the actuator in real time to optimize the sorting trajectory of items.

[0055] According to a third aspect of the embodiments of the present invention,

[0056] An electronic device is provided, comprising:

[0057] processor;

[0058] a memory for storing processor-executable instructions;

[0059] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0060] According to a fourth aspect of the embodiments of the present invention,

[0061] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0062] The fuzzy classification method of items based on composite membership function proposed in this paper constructs a nonlinear mapping relationship between item attributes and fuzzy classification by dynamically dividing attribute subintervals, selecting multiple basic membership functions, parameter adaptive estimation and smooth transition, and effectively improves the accuracy and robustness of item classification. This method fully considers the complexity and diversity of item attribute distribution, and realizes the fine characterization and classification reasoning of item fuzzy features through the design of fuzzy sets and membership functions. In actual item sorting applications, this method can flexibly adapt to the attribute characteristics of different types of items, optimize the classification decision rules by adjusting the definition of fuzzy sets and membership functions, and thus improve the flexibility and intelligence level of the automatic sorting system.

[0063] According to the real-time status of the sorting line and the results of item classification, this paper designs a fuzzy controller to dynamically adjust the key scheduling parameters of the sorting line, constructs a closed-loop feedback control system, and realizes the intelligent scheduling and adaptive optimization of the sorting line. In practical applications, according to the characteristics and needs of specific problems, the input and output variables of the fuzzy controller can be reasonably selected, the design of the fuzzy control rule base and the membership function can be optimized, and the control accuracy and response speed of the fuzzy controller can be improved. At the same time, fuzzy control can also be combined with other modern control theories, such as adaptive control and predictive control, to further enhance the intelligence level and control performance of the sorting line scheduling system, and provide strong support for the realization of efficient, flexible and intelligent logistics sorting process.

[0064] This paper selects appropriate actuators to realize the sorting action of items, establishes the kinematic and dynamic models of the actuators, determines the optimization objectives and constraints of the sorting trajectory, adopts fuzzy PID control and fuzzy sliding mode control algorithms, generates the optimal sorting trajectory offline, and realizes real-time tracking and feedback control of the sorting trajectory online, effectively improving the efficiency, accuracy and flexibility of item sorting. In practical applications, according to the characteristics and requirements of specific problems, the type and structure of the actuator can be reasonably selected, the design parameters and rule base of the fuzzy controller can be optimized, and efficient and reliable trajectory optimization algorithms and control strategies can be developed to realize the automatic planning and precise execution of complex sorting tasks, providing key technical support for the construction and development of intelligent logistics systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a flow chart of a logistics classification and sorting control method based on fuzzy control according to an embodiment of the present invention;

[0066] Figure 2 It is a structural schematic diagram of a logistics classification and sorting control system based on fuzzy control according to an embodiment of the present invention. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0068] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0069] Figure 1 FIG. 1 is a flow chart of a logistics classification and sorting control method based on fuzzy control according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0070] S1. Obtain multi-dimensional attribute information of the items to be sorted, collect attribute data of the items through a multi-sensor array set at different positions of the sorting line, and pre-process the collected attribute data to obtain a standardized item attribute feature vector;

[0071] S2. Based on the feature vector of item attributes, construct fuzzy sets describing different sorting categories, construct a composite membership function for each fuzzy set of item attribute dimensions, and realize the continuity of the composite membership function at the boundary of adjacent sub-intervals through smooth transition technology; comprehensively consider the statistical characteristics of the distribution density and discrete degree of item attribute data points in each sub-interval, adaptively determine the basic function type and corresponding parameters of the composite membership function in each sub-interval, and establish a fuzzy mapping relationship between item attributes and sorting categories; input the multi-dimensional attribute information of the item into the composite membership function of the corresponding attribute dimension, calculate the membership value of the item on each fuzzy set, construct a fuzzy classification rule base, and use fuzzy reasoning method to comprehensively evaluate the membership value of the item to each candidate classification; according to the membership degree of the item to each candidate classification, determine the final classification result of the item through defuzzification processing, and determine the best sorting category of the item;

[0072] S3. According to the optimal sorting category of items, a complete fuzzy control rule base is constructed, and the Mamdani reasoning and center of gravity defuzzification method are used to implement the software algorithm of the fuzzy controller; the actuator is selected to realize the sorting action of items, and according to the control command signal output by the fuzzy controller and the physical parameter model of the actuator, the fuzzy PID control and fuzzy sliding mode control algorithms are used to adjust the position, speed and acceleration of the actuator in real time to optimize the item sorting trajectory.

[0073] In an optional embodiment,

[0074] Based on the item attribute feature vector, a fuzzy set describing different sorting categories is constructed. For the fuzzy set of each item attribute dimension, a composite membership function is constructed. The steps of achieving the continuity of the composite membership function at the boundary of adjacent sub-intervals through smooth transition technology include:

[0075] Obtain multidimensional attribute data of the items to be sorted, and divide the attribute dimension into several non-intersecting sub-intervals according to the value range and data distribution pattern of each attribute dimension, each sub-interval corresponding to a fuzzy set; wherein the division of sub-intervals comprehensively considers the distribution density and discrete degree characteristics of the attribute data in this dimension;

[0076] For each fuzzy set of each attribute dimension, one or more function types are selected from a preset candidate basic membership function library as the basic membership function of the subinterval according to the clustering characteristics of the data points in the corresponding subinterval; the candidate basic membership function library includes a triangular function, a trapezoidal function and a Gaussian function;

[0077] By using the least square method and gradient descent parameter estimation method, the attribute data samples in the subinterval are used as the training set, and the optimal parameter set of the selected basic membership function is adaptively estimated by optimizing the objective function, so as to minimize the fitting error of the data distribution in the subinterval.

[0078] The basic membership functions of all sub-intervals of each attribute dimension are combined and integrated to construct a composite membership function on the attribute dimension. The basic membership functions of adjacent sub-intervals are locally adjusted and optimized by using the smooth transition technology of interpolation fitting and spline function approximation.

[0079] The multi-dimensional attribute feature vector of the item is input into the composite membership function of the corresponding attribute dimension, the membership value of the item on each fuzzy set is calculated, and the nonlinear mapping relationship between the item attribute and the fuzzy classification is constructed.

[0080] For example, in an automatic sorting system, the attribute characteristics of items often present a complex nonlinear distribution, and a single type of membership function is difficult to accurately describe the fuzzy relationship between item attributes and sorting categories. In order to improve the accuracy and adaptability of item classification, this paper proposes an item fuzzy classification method based on a composite membership function. By constructing fuzzy sets that describe different sorting categories, a composite membership function is designed for each item attribute dimension to achieve a nonlinear mapping between item attributes and fuzzy classification. The specific steps are as follows:

[0081] Obtain multi-dimensional attribute data of the items to be sorted. The size, weight, color and other attribute information of the items are collected through sensors and visual systems to form item attribute feature vectors. Taking the sorting of daily necessities in an e-commerce warehouse as an example, the item attributes include five dimensions: length, width, height, weight and color saturation. The length ranges from 5cm to 50cm, the width ranges from 3cm to 30cm, the height ranges from 2cm to 25cm, the weight ranges from 10g to 1000g, and the color saturation ranges from 0 to 100.

[0082] According to the value range and data distribution pattern of each attribute dimension, the attribute dimension is divided into several non-intersecting sub-intervals. Taking into account the distribution density and discreteness of the attribute data on this dimension, a dynamic clustering algorithm, such as K-means clustering or density clustering, is used to adaptively determine the number and boundaries of the sub-intervals. Taking the length dimension as an example, according to the distribution characteristics of historical data, it is divided into three sub-intervals: 5cm to 20cm, 20cm to 35cm, and 35cm to 50cm, corresponding to the fuzzy sets of small items, medium items, and large items, respectively.

[0083] For each fuzzy set of each attribute dimension, one or more function types are selected from the preset candidate basic membership function library according to the clustering characteristics of the data points in the corresponding subinterval as the basic membership function of the subinterval. The candidate function library includes commonly used membership function forms such as triangular function, trapezoidal function and Gaussian function. Taking the fuzzy set of medium-sized items in the length dimension as an example, according to the distribution form of the data points in the subinterval, the trapezoidal function and Gaussian function are selected as the candidate basic membership functions.

[0084] Using the parameter estimation method, the attribute data samples in the subinterval are used as the training set. By optimizing the objective function, the optimal parameter set of the selected basic membership function is adaptively estimated. The objective function usually selects the mean square error or cross entropy loss function to characterize the degree of fit of the membership function to the data distribution. Taking the fuzzy set of medium-sized items in the length dimension as an example, assuming that the selected basic membership function is a Gaussian function, the data samples are fitted by the least squares method, and the optimal mean of the Gaussian function is 27.5cm and the optimal standard deviation is 3.8cm.

[0085] The basic membership functions of all subintervals of each attribute dimension are combined and integrated to construct a composite membership function on the attribute dimension. Smooth transition techniques, such as interpolation fitting or spline function approximation, are used to locally adjust and optimize the basic membership functions of adjacent subintervals to ensure that the composite membership function meets the continuity condition within its domain. Taking the length dimension as an example, assuming that the small item subinterval uses a triangular function, the medium item subinterval uses a Gaussian function, and the large item subinterval uses a trapezoidal function, the parameters of the basic membership function are smoothly transitioned at the subinterval boundary through the piecewise Three-order Hermite interpolation method to obtain a composite membership function on the length dimension.

[0086] The multi-dimensional attribute feature vector of the item is input into the composite membership function of the corresponding attribute dimension, and the membership value of the item on each fuzzy set is calculated. Taking an item with a length of 24 cm, a width of 15 cm, a height of 10 cm, a weight of 300 g, and a color saturation of 80 as an example, its attribute feature vector is substituted into the composite membership function, and its membership on each fuzzy set is obtained as follows: in the length dimension, the membership of small items is 0.2, the membership of medium items is 0.8, and the membership of large items is 0; in the width dimension, the membership of small items is 0.1, and the membership of medium items is 0.9; in the height dimension, the membership of small items is 0.6, and the membership of medium items is 0.4; in the weight dimension, the membership of light items is 0.3, and the membership of medium items is 0.7; in the color saturation dimension, the membership of low-saturation items is 0.1, the membership of medium-saturation items is 0.2, and the membership of high-saturation items is 0.7.

[0087] According to the membership values ​​of the items in each fuzzy set, the fuzzy logic reasoning rules are comprehensively applied to determine the final sorting category of the items. Commonly used fuzzy logic reasoning rules include the maximum membership method, the weighted average method, and the fuzzy matrix synthesis method. Taking the above-mentioned items as an example, according to the maximum membership method, the items belong to the fuzzy sets of medium-sized items, medium-sized items, medium-sized items, and high-saturation items in the length, width, weight, and color saturation dimensions, respectively, and belong to the fuzzy set of small items in the height dimension. Considering the importance weight of each attribute dimension, the item is finally classified as a medium-sized item and transported to the corresponding sorting channel.

[0088] In summary, the fuzzy classification method of items based on composite membership function proposed in this paper constructs a nonlinear mapping relationship between item attributes and fuzzy classification by dynamically dividing attribute subintervals, selecting multiple basic membership functions, parameter adaptive estimation and smooth transition, and effectively improves the accuracy and robustness of item classification. This method fully considers the complexity and diversity of item attribute distribution, and realizes the fine characterization and classification reasoning of item fuzzy features through the design of fuzzy sets and membership functions. In actual item sorting applications, this method can flexibly adapt to the attribute characteristics of different types of items, and optimize the classification decision rules by adjusting the definition of fuzzy sets and membership functions, thereby improving the flexibility and intelligence level of the automatic sorting system.

[0089] In an optional embodiment,

[0090] The steps of using the least squares method and gradient descent parameter estimation method, taking the attribute data samples in the subinterval as the training set, and adaptively estimating the optimal parameter set of the selected basic membership function by optimizing the objective function so as to minimize the fitting error of the data distribution in the subinterval include:

[0091] For a certain attribute dimension of the item to be classified, all attribute data points in each subinterval after the attribute dimension is divided are obtained, and the attribute data points are used as a training sample set, where each sample point includes the attribute value of the point and a preset true membership value;

[0092] According to the type of the selected basic membership function, an optimization objective function is constructed with the goal of minimizing the sum of squares of fitting errors. The objective function is used to measure the degree of fit of the basic membership function to the sample points in the subinterval.

[0093] Initialize the parameter vector of the basic membership function. For the triangular function and trapezoidal function, the parameter vector includes the left and right boundaries and the peak point. For the Gaussian function, the parameter vector includes the mean and standard deviation.

[0094] The constructed objective function is iteratively optimized using the gradient descent method. The gradient vector of the objective function with respect to each parameter is calculated through partial derivatives, and the parameters are updated along the negative gradient direction until the objective function value converges or the maximum number of iterations is reached.

[0095] The parameter vector at convergence after iterative optimization is used as the optimal parameter estimate of the basic membership function in this subinterval, so as to minimize the fitting error of the attribute data distribution in this subinterval and obtain the basic membership function after parameter optimization.

[0096] For example, in order to make the composite membership function accurately describe the relationship between item attributes and fuzzy classification, it is necessary to optimize the parameters of the basic membership function selected in each subinterval so as to minimize the fitting error of the attribute data distribution in the subinterval. This paper adopts a parameter adaptive estimation method based on the gradient descent method to obtain the optimal parameter set of the basic membership function by iteratively optimizing the objective function. The specific steps are as follows:

[0097] For a certain attribute dimension of the item to be classified, all attribute data points in each sub-interval after the attribute dimension is divided are obtained, and the attribute data points are used as the training sample set. Each sample point includes the attribute value of the point and the preset true membership value. Taking the sub-interval of medium-sized items in the length dimension as an example, assuming that the sub-interval ranges from 20cm to 35cm, 100 sample points are selected from the historical data, and the attribute value of one sample point is 28cm, and the corresponding true membership value is 0.85.

[0098] According to the type of basic membership function selected, an optimization objective function is constructed with the goal of minimizing the sum of squares of fitting errors. The objective function is used to measure the degree of fit of the basic membership function to the sample points in the subinterval. Its expression is the sum of squares of fitting errors. The fitting error is defined as the absolute value of the difference between the true membership value of the sample point and the predicted membership value calculated by the basic membership function. Taking the Gaussian function as an example, its parameter vector includes the mean μ and the standard deviation σ. The optimization objective function can be expressed as the sum of squares of fitting errors of all sample points, that is, Σ(y i - exp(-((x i -μ) / σ) 2 )) 2 , where x i is the attribute value of the i-th sample point, y i is the true membership value of the i-th sample point.

[0099] Initialize the parameter vector of the basic membership function. For the triangular function and trapezoidal function, the parameter vector includes the left and right boundaries and the peak point, and the initial value can be set according to the boundary and midpoint of the sub-interval; for the Gaussian function, the parameter vector includes the mean μ and the standard deviation σ, and the initial value can be set according to the midpoint and width of the sub-interval. Taking the Gaussian function as an example, assuming that the sub-interval range is 20cm to 35cm, the initial value of μ can be set to 27.5cm, and the initial value of σ can be set to 3.75cm.

[0100] The constructed objective function is iteratively optimized using the gradient descent method. First, the partial derivatives of the objective function with respect to each parameter are calculated to obtain the gradient vector; then, the parameters are updated along the negative gradient direction, that is, the new parameter value is equal to the current parameter value minus the gradient vector multiplied by the learning rate; the iterative update process is repeated until the objective function value converges or the maximum number of iterations is reached. Taking the Gaussian function as an example, assuming that the learning rate is 0.01 and the maximum number of iterations is 1000, in each iteration, the partial derivatives of the objective function with respect to μ and σ are calculated respectively, and μ and σ are updated according to the values ​​of the partial derivatives until the convergence condition is met or the number of iterations reaches 1000.

[0101] The parameter vector when iterative optimization converges is used as the optimal parameter estimate of the basic membership function in the subinterval. The basic membership function is updated using the optimal parameters to obtain the membership function expression that minimizes the fitting error of the attribute data distribution in the subinterval.

[0102] In summary, by introducing the parameter adaptive estimation step based on the gradient descent method, the parameter values ​​of the basic membership function selected in each subinterval can be automatically optimized to minimize the fitting error of the attribute data distribution in the subinterval. This step can significantly improve the accuracy and adaptability of the composite membership function and provide more reliable membership calculation results for subsequent fuzzy classification decisions. In practical applications, the appropriate basic membership function type and parameter estimation method can be flexibly selected according to the characteristics of the specific problem, such as considering introducing a regularization term to prevent overfitting, or using a heuristic optimization algorithm to accelerate the convergence process, to further improve the performance and efficiency of the membership function.

[0103] In an optional embodiment,

[0104] The steps of combining and integrating the basic membership functions of all subintervals of each attribute dimension to construct a composite membership function on the attribute dimension, and using the smooth transition technology of interpolation fitting and spline function approximation to locally adjust and optimize the basic membership functions of adjacent subintervals include:

[0105] According to the order of sub-intervals, the basic membership functions after parameter optimization on each sub-interval are sequentially spliced ​​to form a piecewise function covering the entire attribute value range;

[0106] For the intersection points of adjacent sub-intervals, a smooth transition technology of interpolation fitting or spline function approximation is used to construct a transition function that meets the continuity and smoothness conditions;

[0107] The smoothed piecewise function and transition function are integrated to construct a composite membership function covering the entire attribute dimension, so that it is consistent with the optimized basic membership function within the sub-interval and achieves continuous and smooth transition at the junction of adjacent sub-intervals.

[0108] On the premise of maintaining continuity and smoothness, the local shape parameters of the composite membership function are fine-tuned, an additional correction function is introduced, appropriate regularization constraints are set, and numerical optimization methods are used to locally adjust and optimize the composite membership function.

[0109] For example, after completing the parameter optimization of the basic membership function in each sub-interval, it is necessary to combine and integrate these local membership functions to construct a composite membership function covering the entire attribute dimension. In order to ensure that the composite membership function meets the continuity and smoothness conditions within its definition domain, it is necessary to use smooth transition techniques such as interpolation fitting and spline function approximation to locally adjust and optimize the basic membership functions of adjacent sub-intervals. The specific steps are as follows:

[0110] According to the order of the sub-intervals, the basic membership functions after parameter optimization on each sub-interval are spliced ​​in sequence to form a piecewise function covering the entire attribute value range. Ensure that the piecewise function is consistent with the corresponding basic membership function in each sub-interval. Taking the length dimension as an example, assuming that the dimension is divided into three sub-intervals of small items, medium items, and large items, and the corresponding optimized basic membership functions are small functions, medium functions, and large functions, respectively, the piecewise function can be expressed as: when the attribute value is between 5 and 20, the membership function takes the value of the small function; when the attribute value is between 20 and 35, the membership function takes the value of the medium function; when the attribute value is between 35 and 50, the membership function takes the value of the large function.

[0111] For the intersection of adjacent sub-intervals, a smooth transition technique using interpolation fitting or spline function approximation is used to construct a transition function that meets the continuity and smoothness conditions, so as to achieve seamless connection of the membership functions of adjacent sub-intervals at the intersection. Commonly used smooth transition techniques include linear interpolation, cubic Hermite interpolation, B-spline curve fitting, etc. Taking the intersection of the small and medium-sized item sub-intervals in the length dimension as an example, assuming that the attribute value of the intersection point is 20, and linear interpolation is used to construct the transition function, the transition function can be expressed as follows: in the interval of attribute values ​​20 to 22, the value of the membership function linearly transitions from the function value of the small function at the intersection point to the function value of the medium function at the intersection point, and the width of the transition interval is 2.

[0112] The smoothed piecewise function and transition function are integrated to construct a composite membership function covering the entire attribute dimension. The composite membership function is made consistent with the optimized basic membership function within the subinterval, and a continuous smooth transition is achieved at the junction of adjacent subintervals. Taking the length dimension as an example, assuming that the transition function at the junction of the small and medium-sized subintervals is the first transition function, and the transition function at the junction of the medium and large subintervals is the second transition function, the composite membership function can be expressed as follows: when the attribute value is between 5 and 20, the membership function takes the value of the small function; when the attribute value is between 20 and 22, the membership function takes the value of the first transition function; when the attribute value is between 22 and 35, the membership function takes the value of the medium function; when the attribute value is between 35 and 37, the membership function takes the value of the second transition function; when the attribute value is between 37 and 50, the membership function takes the value of the large function.

[0113] Under the premise of maintaining continuity and smoothness, the local shape parameters of the composite membership function are fine-tuned, additional correction functions are introduced, appropriate regularization constraints are set, and numerical optimization methods are used to locally adjust and optimize the composite membership function. This step aims to improve the expressiveness and flexibility of the composite membership function so that it can better adapt to the data distribution characteristics in different sub-intervals. Taking the medium-sized item sub-interval in the length dimension as an example, assuming that a quadratic correction function is introduced in this sub-interval and the membership function of this sub-interval is locally adjusted, the adjusted composite membership function can be expressed as follows: In the medium-sized item sub-interval, the value of the membership function is obtained by multiplying the value of the original composite membership function by the value of the quadratic correction function and a weight coefficient, where the weight coefficient and the parameters of the quadratic correction function can be adaptively estimated according to the training sample data in the sub-interval using methods such as least squares or cross-validation, so that the adjusted composite membership function has higher fitting accuracy and flexibility in the sub-interval.

[0114] Through the above steps, the optimized basic membership functions on each subinterval can be organically combined into a continuous and smooth composite membership function, which not only ensures the consistency of the membership function with the basic function in the local subinterval, but also achieves smooth transition and good continuity between adjacent subintervals. In addition, by introducing additional correction functions and regularization constraints, the expressiveness and adaptability of the composite membership function can be further improved, so that it can better characterize the complex relationship between item attributes and fuzzy classification. In practical applications, according to the characteristics and needs of specific problems, appropriate smooth transition technology and local adjustment methods can be flexibly selected, and the performance of the composite membership function can be evaluated and optimized through cross-validation and other means, in order to obtain the best fuzzy classification effect.

[0115] In an optional embodiment,

[0116] The steps of inputting the multi-dimensional attribute information of the item into the composite membership function of the corresponding attribute dimension, calculating the membership value of the item on each fuzzy set, building a fuzzy classification rule base, and using the fuzzy reasoning method to comprehensively evaluate the membership value of the item to each candidate classification include:

[0117] The multi-dimensional attribute information of the items to be sorted is extracted and input into the composite membership function of the corresponding attribute dimension. The membership value of the fuzzy set corresponding to each attribute dimension of the item is calculated by the composite membership function. The membership value indicates the degree of belonging of the item to the fuzzy set corresponding to the attribute dimension.

[0118] Construct a fuzzy classification rule library. The fuzzy classification rule is composed of the combination conditions of the attribute membership values ​​of the items as the antecedent and the classification results of the items as the consequent, which is used to describe the fuzzy mapping relationship between the item attributes and the classification categories.

[0119] For each fuzzy classification rule, the attribute membership value of the item is substituted into the antecedent part of the fuzzy classification rule, and the trigger strength of each fuzzy classification rule is calculated by the fuzzy reasoning method. The trigger strength is calculated by the combination of the attribute membership values ​​of each item, which indicates the degree to which the fuzzy classification rule is satisfied by the item attributes.

[0120] For each candidate classification category, summarize the reasoning results of all fuzzy classification rules pointing to this category, and take the maximum trigger strength of the item in all fuzzy classification rules pointing to this category as the item's membership degree to this category. The membership degree represents the belonging measure of the item to this classification category based on fuzzy reasoning.

[0121] According to the degree of belonging of the items to each candidate classification category, a decision strategy is adopted to determine the final classification of the items, and the items are divided into the classification category with the highest degree of belonging.

[0122] For example, after constructing the composite membership function on each attribute dimension, the multi-dimensional attribute information of the item can be input into the corresponding membership function, the membership value of the item on each fuzzy set can be calculated, and the fuzzy classification rule base can be combined to use the fuzzy reasoning method to comprehensively evaluate the membership value of the item, and finally determine the classification of the item. The specific steps are as follows:

[0123] The multi-dimensional attribute information of the items to be sorted is extracted and input into the composite membership function of the corresponding attribute dimension. The membership value of the corresponding fuzzy set of the item in each attribute dimension is calculated through the composite membership function. The membership value indicates the degree of belonging of the item to the corresponding fuzzy set in the attribute dimension. Taking the three attribute dimensions of length, width and height as an example, assuming that the attribute values ​​of an item are 30cm, 20cm and 15cm respectively, these attribute values ​​are input into the composite membership function of the corresponding dimension, and it can be obtained that the item belongs to the "medium" fuzzy set in the length dimension with a membership of 0.8, the "moderate" fuzzy set in the width dimension with a membership of 0.6, and the "short" fuzzy set in the height dimension with a membership of 0.7.

[0124] Construct a fuzzy classification rule base. The fuzzy classification rules are composed of the combination of the attribute membership values ​​of the items as the antecedent and the classification results of the items as the consequent, which are used to describe the fuzzy mapping relationship between the item attributes and the classification categories. Taking parcel sorting as an example, the following fuzzy classification rules can be constructed: if the length of the item is "medium", the width is "moderate", and the height is "short", then the item belongs to the "medium package" category; if the length of the item is "long", the width is "wide", and the height is "tall", then the item belongs to the "large package" category; if the length of the item is "short", the width is "narrow", and the height is "short", then the item belongs to the "small package" category.

[0125] For each fuzzy classification rule, the attribute membership value of the item is substituted into the antecedent of the fuzzy classification rule, and the trigger strength of each fuzzy classification rule is calculated by the fuzzy reasoning method. The trigger strength is calculated by the combination of the attribute membership values ​​of each item, which indicates the degree to which the fuzzy classification rule is satisfied by the item attributes. Commonly used fuzzy reasoning methods include Mamdani reasoning, Larsen reasoning, and reasoning based on correlation.

[0126] For each candidate classification category, the reasoning results of all fuzzy classification rules pointing to this category are summarized, and the maximum trigger intensity of the item in all fuzzy classification rules pointing to this category is taken as the item's degree of membership to this category. The degree of membership represents the measure of the item's belonging to this classification category based on fuzzy reasoning.

[0127] According to the degree of membership of the item to each candidate classification category, the decision strategy is used to determine the final classification of the item, and the item is divided into the classification category with the largest degree of membership, so as to realize the fuzzy judgment and classification of the item category. Commonly used decision strategies include the maximum membership method, the weighted average method, the centroid method, etc.

[0128] Through the above steps, the multi-dimensional attribute information of the items can be combined with the fuzzy classification rule base, and the fuzzy reasoning method can be used to calculate the degree of membership of the items to each candidate category, and the final classification of the items can be determined according to the degree of membership, so as to realize the fuzzy classification and intelligent judgment of the items. In practical applications, according to the characteristics and needs of specific problems, the fuzzy classification rule base can be reasonably designed, and appropriate fuzzy reasoning methods and judgment strategies can be selected. By optimizing the coverage and accuracy of the fuzzy classification rule base, the accuracy and efficiency of item classification can be improved. At the same time, an adaptive learning mechanism can be introduced to dynamically adjust and optimize the fuzzy classification rule base according to the feedback results of item classification, so that it can adapt to changes in the distribution of item attributes and improve the robustness and intelligence level of the classification system.

[0129] In an optional embodiment,

[0130] According to the best sorting category of items, a complete set of fuzzy control rule base is constructed, and the steps of implementing the software algorithm of the fuzzy controller using Mamdani reasoning and center of gravity defuzzification method include:

[0131] Input the input variable value of the current control cycle into the fuzzy controller to trigger the corresponding fuzzy control rule. According to the input variable language value combination of the antecedent part of each fuzzy control rule, calculate the membership value of each input variable corresponding to the language value, and use the minimum operation to obtain the triggering strength of each fuzzy control rule.

[0132] According to the triggering strength of each fuzzy control rule, the membership function of the output variable of the consequent part of the fuzzy control rule is truncated to obtain the membership function of the output variable under each fuzzy control rule;

[0133] The membership functions of all output variables that trigger the fuzzy control rules are combined, and the total membership function of the output variables is obtained by taking the largest operation;

[0134] The centroid defuzzification method is used to convert the total membership function of the output variable into an accurate control quantity. The centroid defuzzification obtains the defuzzified value by calculating the center of the area under the total membership function. In the actual calculation, the total membership function is discretized into a finite number of points by numerical integration, and then the centroid is approximately calculated by weighted average.

[0135] The numerical value obtained by defuzzifying the center of gravity is converted into a specific actuator control instruction, and the control instruction is sent to the actuator.

[0136] For example, after determining the best sorting category for items, it is necessary to design a fuzzy controller to adjust the scheduling parameters of the sorting line in real time according to the real-time operating status of the sorting line and the dynamic constraints of the actuator, so as to optimize the operating efficiency and sorting quality of the sorting line. The specific steps are as follows:

[0137] Design the input and output language variables of the fuzzy controller. The input variables may include the current running speed of the sorting line, the arrival interval of the items, the state of the actuator, etc. The output variables may include the speed adjustment of the sorting line, the action control parameters of the actuator, etc. Taking the speed control of the sorting line as an example, the input variable "current speed" can be divided into three language values: "low speed", "medium speed", and "high speed", and the output variable "speed adjustment" can be divided into three language values: "deceleration", "maintain", and "acceleration".

[0138] Construct a complete set of fuzzy control rule bases. The fuzzy control rules are composed of the language value combination of the input variables as the antecedent and the language value of the output variable as the consequent. Taking the sorting line speed control as an example, the following fuzzy control rules can be constructed: if the current speed is "low speed" and the item interval time is "long", the speed adjustment amount is "acceleration"; if the current speed is "high speed" and the item interval time is "short", the speed adjustment amount is "deceleration"; if the current speed is "medium speed" and the item interval time is "moderate", the speed adjustment amount is "maintain".

[0139] Input the input variable value of the current control cycle into the fuzzy controller to trigger the corresponding fuzzy control rule. According to the input variable language value combination of the antecedent part of each fuzzy control rule, calculate the membership value of each input variable corresponding to the language value, and use the smaller operation to obtain the triggering strength of each fuzzy control rule. Taking a certain control cycle as an example, assuming that the current sorting line speed is 2m / s, the membership value corresponding to the "medium speed" language value is 0.8, and the membership value corresponding to the "high speed" language value is 0.2; the item interval time is 3s, the membership value corresponding to the "moderate" language value is 0.6, and the membership value corresponding to the "long" language value is 0.4. Then the strength of triggering the second fuzzy control rule is min(0.8, 0.6) = 0.6, and the strength of triggering the third fuzzy control rule is min(0.2, 0.4) = 0.2.

[0140] According to the triggering strength of each fuzzy control rule, the membership function of the output variable of the consequent part of the fuzzy control rule is truncated to obtain the membership function of the output variable under each fuzzy control rule. Taking the two triggered fuzzy control rules as an example, it is assumed that the membership function corresponding to the language value "maintain" of the output variable "speed adjustment amount" is a triangular membership function with a peak value at 0 and a support interval of -0.2 to 0.2; the membership function corresponding to the language value "deceleration" is a Z-type membership function with a turning point at -0.3. Under the second fuzzy control rule, the membership function of the output variable is the shape of the "maintain" membership function truncated to a height of 0.6; under the third fuzzy control rule, the membership function of the output variable is the shape of the "deceleration" membership function truncated to a height of 0.2.

[0141] All output variable membership functions that trigger fuzzy control rules are merged, and the total membership function of the output variables is obtained by taking the maximum operation. Taking the two output membership functions obtained above as an example, for each value of the output variable, the maximum membership value of the two membership functions at that value is taken as the membership value of the total membership function at that value, and a new merged membership function curve is obtained, which represents the fuzzy distribution of the output variable in the current control cycle.

[0142] The total membership function of the output variable is converted into an accurate control quantity by using the centroid defuzzification method. The centroid defuzzification obtains the defuzzified value by calculating the center of the area under the total membership function. In actual calculation, the total membership function is discretized into a finite number of points by numerical integration, and then the centroid is approximately calculated by weighted average. Taking the total membership function obtained above as an example, assuming that 50 points are sampled at equal intervals in the domain of the output variable [-0.5, 0.5], the sum of the product of the membership value at each point and the coordinate value of the point is divided by the sum of the membership values ​​of each point, and the centroid coordinate of the output variable is -0.15, indicating that the sorting line speed should be reduced by 0.15m / s.

[0143] The numerical value obtained by defuzzifying the center of gravity is converted into a specific actuator control instruction, and the control instruction is sent to the actuator to realize real-time regulation of the controlled object. Taking the above calculation results as an example, the speed adjustment amount of -0.15m / s is converted into the control instruction of the servo motor, and the actual operating frequency of the sorting line motor is adjusted to reduce the operating speed of the sorting line by 0.15m / s, thereby optimizing the operating state of the sorting line and improving the efficiency and accuracy of item sorting.

[0144] Through the above steps, we can design a fuzzy controller to dynamically adjust the key scheduling parameters of the sorting line according to the real-time status of the sorting line and the results of item classification, build a closed-loop feedback control system, and realize intelligent scheduling and adaptive optimization of the sorting line. In practical applications, we can reasonably select the input and output variables of the fuzzy controller according to the characteristics and needs of specific problems, optimize the design of the fuzzy control rule base and membership function, and improve the control accuracy and response speed of the fuzzy controller. At the same time, we can also combine fuzzy control with other modern control theories, such as adaptive control and predictive control, to further enhance the intelligence level and control performance of the sorting line scheduling system, and provide strong support for realizing efficient, flexible and intelligent logistics sorting processes.

[0145] In an optional embodiment,

[0146] The actuator is selected to realize the sorting action of the items. According to the control command signal output by the fuzzy controller and the physical parameter model of the actuator, the fuzzy PID control and fuzzy sliding mode control algorithms are used to adjust the position, speed and acceleration of the actuator in real time. The steps of optimizing the item sorting trajectory include:

[0147] Establishing a kinematic model and a dynamic model of the actuator, establishing the forward and inverse kinematic equations of the actuator by using the DH parameter method, describing the mapping relationship between the joint angle and the end position and posture, and considering the dynamic characteristics of the actuator to establish the dynamic equation of the actuator;

[0148] Determine the optimization objective function and constraint conditions of the sorting trajectory, wherein the optimization objective function includes minimizing the sorting time, maximizing the average speed and minimizing the acceleration change, and the constraint conditions include the upper and lower limits of the joint angle, angular velocity and angular acceleration of the actuator and the boundary constraint of the end position; discretize the sorting trajectory, generate smooth trajectory segments between adjacent discrete nodes by interpolation method, and transform the continuous trajectory planning problem into a discrete optimization problem;

[0149] A fuzzy PID control algorithm is applied, with position error and speed error as input, and the parameters of the PID controller are adjusted online through fuzzy reasoning and defuzzification calculation to generate optimized position and speed control instructions, and the optimized position and speed instructions are substituted into the kinematic model of the actuator to calculate the angle and angular velocity of each joint; a fuzzy sliding mode control algorithm is applied, with acceleration error and acceleration change rate as input, and the sliding mode surface and control law are adjusted online through fuzzy reasoning and defuzzification calculation to generate optimized acceleration control instructions, and the optimized acceleration instructions are substituted into the dynamic model of the actuator to calculate the driving torque of each joint;

[0150] The sorting trajectory is iteratively optimized, and the position, velocity and acceleration instructions obtained by the fuzzy PID control algorithm and the fuzzy sliding mode control algorithm are used as the optimal values ​​of the current discrete node, substituted into the next discrete node, and the fuzzy PID control and fuzzy sliding mode control are repeatedly executed until the optimization of the entire sorting trajectory is completed.

[0151] For example, after obtaining the best sorting category of items and the optimized scheduling parameters of the sorting line, it is necessary to select a suitable actuator to implement the sorting action of the items, and optimize the item sorting trajectory according to the control command signal output by the fuzzy controller and the physical parameter model of the actuator to improve the sorting efficiency and accuracy. The specific steps are as follows:

[0152] Establish the kinematic model and dynamic model of the actuator. Take the three-degree-of-freedom serial robot arm as an example, assume that its joint angles are joint 1 angle, joint 2 angle, and joint 3 angle, and the end positions are the end X coordinate, end Y coordinate, and end Z coordinate. Through the DH parameter table and the homogeneous transformation matrix, establish the positive kinematic equation that describes the mapping relationship between the joint angle and the end position and posture. At the same time, considering the mass, moment of inertia and other parameters of the robot arm, establish the dynamic equation that describes the force of the robot arm, where the left side of the equation is the sum of the inertia force, centripetal force, Coriolis force and gravity of the robot arm, and the right side of the equation is the joint driving torque.

[0153] Determine the optimization objective function and constraints of the sorting trajectory. Taking the three-degree-of-freedom serial robot as an example, the optimization objectives are to minimize the sorting time, maximize the average speed, and minimize the acceleration change. The constraints are: the angle of joint 1 is limited to between -90 and 90 degrees, the angle of joint 2 is limited to between -90 and 90 degrees, and the angle of joint 3 is limited to between -90 and 90 degrees; the angular velocity of joint 1 is limited to between -45 and 45 degrees per second, the angular velocity of joint 2 is limited to between -45 and 45 degrees per second, and the angular velocity of joint 3 is limited to between -45 and 45 degrees per second; the angular acceleration of joint 1 is limited to between -22.5 and 22.5 degrees per square second, the angular acceleration of joint 2 is limited to between -22.5 and 22.5 degrees per square second, and the angular acceleration of joint 3 is limited to between -22.5 and 22.5 degrees per square second; the end position is limited to inside a sphere with a radius of 0.5 meters and the sorting point as the center.

[0154] Discretize the sorting trajectory. Taking the straight line trajectory as an example, set the starting coordinates to (0.2m, 0.3m, 0.4m) and the end coordinates to (0.7m, 0.6m, 0.5m), divide the trajectory into 10 discrete nodes at equal distances, and generate a smooth curve as a trajectory segment through cubic spline interpolation between every two adjacent nodes. Then the entire sorting trajectory can be represented as the connection of 10 trajectory segments. The position, velocity, and acceleration on each trajectory segment are determined by the interpolation function, and the optimization variables are the position coordinates of the 11 discrete nodes.

[0155] The fuzzy PID control algorithm is applied, with position error and speed error as input. Through fuzzy reasoning and defuzzification calculation, the proportional coefficient, integral coefficient and differential coefficient of the PID controller are adjusted online to generate the optimized position control quantity and speed control quantity, and then substituted into the kinematic model of the actuator to calculate the angle and angular velocity of each joint. Taking the three-degree-of-freedom serial robot as an example, the current end position is (0.45m, 0.38m, 0.42m), the expected position is (0.5m, 0.4m, 0.45m), the current end speed is (0.1m per second, 0.2m per second, 0.15m per second), the expected speed is (0.2m per second, 0.3m per second, 0.25m per second), the position error and speed error are (-0.05m, -0.02m, -0.03m) and (-0.1m per second, -0.1m per second, -0.1m per second). The error is input into the fuzzy PID controller. Through reasoning and defuzzification, the optimized PID parameters are obtained as proportional coefficient 5.2, integral coefficient 0.5, differential coefficient 1.8. Substitute them into the PID control formula to calculate the position control amount (0.52 meters, 0.48 meters, 0.54 meters) and speed control amount (0.24 meters per second, 0.32 meters per second, 0.28 meters per second). Substitute the control amount into the inverse kinematics equation of the robot arm and calculate the corresponding joint angles of 21.8 degrees, 44.2 degrees, and -14.9 degrees, respectively, and the joint angular velocities of 17.8 degrees per second, 25.8 degrees per second, and 10.3 degrees per second, respectively.

[0156] The fuzzy sliding mode control algorithm is applied, with acceleration error and acceleration change rate as input. Through fuzzy reasoning and defuzzification calculation, the sliding mode function and control law are adjusted online to generate the optimized acceleration control quantity, and then substituted into the dynamic model of the actuator to calculate the driving torque of each joint. Taking the three-degree-of-freedom serial robot as an example, the current end acceleration is (0.5 meters per square second, 0.4 meters per square second, 0.6 meters per square second), the expected acceleration is (0.8 meters per square second, 0.7 meters per square second, 0.9 meters per square second), the acceleration change rate is (1.2 meters per cubic second, 1.5 meters per cubic second, 1.3 meters per cubic second), and the acceleration error is (-0.3 meters per square second, -0.3 meters per square second, -0.3 meters per square second). The error and rate of change are input into the fuzzy sliding mode controller. Through reasoning and defuzzification, the optimized sliding mode function is obtained as 0.5 times the acceleration error plus 1.5 times the acceleration change rate. The control law is the negative 1.2 times sliding mode function sign function. Substituting into the sliding mode control formula, the acceleration control amount is calculated to be 0.84 meters per square second, 0.78 meters per square second, and 0.93 meters per square second. Substituting the control amount into the dynamic equation of the robot arm, the corresponding joint drive torque is calculated to be 2.8 Newton meters, 5.3 Newton meters, and 3.6 Newton meters.

[0157] Iterative optimization of sorting trajectory. Taking the straight line trajectory as an example, assume that the current discrete node is the 5th node, and its position coordinates are (0.45m, 0.38m, 0.42m). The optimized position coordinates obtained by fuzzy PID control are (0.52m, 0.48m, 0.54m), and the optimized acceleration obtained by fuzzy sliding mode control is (0.84m / s2, 0.78m / s2, 0.93m / s2). The optimized position coordinates are used as the starting point of the 6th discrete node, and the optimized acceleration is used as the initial acceleration of the 6th node. Repeat the fuzzy PID control algorithm and the fuzzy sliding mode control algorithm until the optimization of the 10th node is completed, and the optimal position, speed and acceleration values ​​of the entire sorting trajectory at the discrete node are obtained.

[0158] The optimal sorting trajectory is generated by interpolating the optimal values ​​of discrete nodes, and the trajectory is converted into control instructions in the joint space. Taking the optimal value of discrete nodes as an example, a continuous and smooth end position, velocity and acceleration curve is generated by cubic spline interpolation as the optimal sorting trajectory. The optimal sorting trajectory is substituted into the inverse kinematics equation of the robot arm to calculate the angle, angular velocity and angular acceleration control instructions of each joint as the input of the actuator.

[0159] During the item sorting process, the state information of the actuator is collected in real time, the tracking error with the optimal sorting trajectory is calculated, and the tracking error is input into the fuzzy controller as a feedback signal to generate corrected control instructions to achieve real-time tracking and feedback control of the sorting trajectory. Taking a three-degree-of-freedom serial robot as an example, suppose that at a certain moment the actual position of the end of the robot is (0.48 m, 0.42 m, 0.51 m), the actual speed is (0.25 m per second, 0.35 m per second, 0.3 m per second), and the actual acceleration is (0.9 m per square second, 0.85 m per square second, 1.1 m per square second). The expected position of the optimal sorting trajectory at this moment is (0.5 m, 0.45 m, 0.55 m), the expected speed is (0.3 m per second, 0.4 m per second, 0.35 m per second), and the expected acceleration is (1.0 m per square second, 0.95 m per square second, 1.2 m per square second). Then the position tracking error is (-0.02 m, -0.03 m, -0.04 m), and the speed tracking error is (-0.05 m per second, -0.05 m per second, The tracking error is (-0.1 m / s, -0.1 m / s, -0.1 m / s). The tracking error is input into the fuzzy controller, and through reasoning and defuzzification, the corrected position control amount (0.51 m, 0.46 m, 0.57 m), speed control amount (0.32 m / s, 0.42 m / s, 0.37 m / s), and acceleration control amount (1.05 m / s, 1.0 m / s, 1.25 m / s) are generated as real-time control instructions for the actuator, so that the end of the robot arm can track the optimal sorting trajectory more accurately at the next moment.

[0160] Through the above steps, we can select appropriate actuators to implement the sorting action of items, establish the kinematic and dynamic models of the actuators, determine the optimization objectives and constraints of the sorting trajectory, use fuzzy PID control and fuzzy sliding mode control algorithms, generate the optimal sorting trajectory offline, and realize real-time tracking and feedback control of the sorting trajectory online, effectively improving the efficiency, accuracy and flexibility of item sorting. In practical applications, we can reasonably select the type and structure of the actuator according to the characteristics and requirements of specific problems, optimize the design parameters and rule base of the fuzzy controller, develop efficient and reliable trajectory optimization algorithms and control strategies, realize automatic planning and precise execution of complex sorting tasks, and provide key technical support for the construction and development of intelligent logistics systems.

[0161] Figure 2 FIG. 1 is a schematic diagram of the structure of a logistics classification and sorting control system based on fuzzy control according to an embodiment of the present invention. Figure 2 As shown, the system comprises:

[0162] The first unit is used to obtain multi-dimensional attribute information of the items to be sorted, collect attribute data of the items through a multi-sensor array set at different positions of the sorting line, and pre-process the collected attribute data to obtain a normalized attribute feature vector of the items;

[0163] The second unit is used to construct fuzzy sets describing different sorting categories based on the feature vector of item attributes, construct a composite membership function for the fuzzy set of each item attribute dimension, and realize the continuity of the composite membership function at the boundary of adjacent sub-intervals through smooth transition technology; comprehensively consider the statistical characteristics of the distribution density and discrete degree of the item attribute data points in each sub-interval, adaptively determine the basic function type and corresponding parameters of the composite membership function in each sub-interval, and establish a fuzzy mapping relationship between item attributes and sorting categories; input the multi-dimensional attribute information of the item into the composite membership function of the corresponding attribute dimension, calculate the membership value of the item on each fuzzy set, construct a fuzzy classification rule base, and use fuzzy reasoning method to comprehensively evaluate the membership value of the item to each candidate classification; according to the membership degree of the item to each candidate classification, determine the final classification result of the item through defuzzification processing, and determine the best sorting category of the item;

[0164] The third unit is used to build a complete fuzzy control rule base according to the optimal sorting category of items, and use Mamdani reasoning and center of gravity defuzzification method to implement the software algorithm of the fuzzy controller; select actuators to realize the sorting action of items, and use fuzzy PID control and fuzzy sliding mode control algorithms according to the control command signal output by the fuzzy controller and the physical parameter model of the actuator to adjust the position, speed and acceleration of the actuator in real time to optimize the sorting trajectory of items.

[0165] According to a third aspect of the embodiments of the present invention,

[0166] An electronic device is provided, comprising:

[0167] processor;

[0168] a memory for storing processor-executable instructions;

[0169] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0170] According to a fourth aspect of the embodiments of the present invention,

[0171] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0172] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0173] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A logistics classification and sorting control method based on fuzzy control, characterized in that: include: Acquire multi-dimensional attribute information of the items to be sorted, collect attribute data of the items through a multi-sensor array set at different positions of the sorting line, and pre-process the collected attribute data to obtain a standardized attribute feature vector of the items; Based on the feature vector of item attributes, a fuzzy set describing different sorting categories is constructed. For the fuzzy set of each item attribute dimension, a composite membership function is constructed, and the continuity of the composite membership function at the boundary of adjacent sub-intervals is achieved through smooth transition technology; the statistical characteristics of the distribution density and discrete degree of the item attribute data points in each sub-interval are comprehensively considered, and the basic function type and corresponding parameters of the composite membership function in each sub-interval are adaptively determined to establish a fuzzy mapping relationship between item attributes and sorting categories; the multi-dimensional attribute information of the item is input into the composite membership function of the corresponding attribute dimension, the membership value of the item on each fuzzy set is calculated, and a fuzzy classification rule base is constructed. The fuzzy reasoning method is used to comprehensively evaluate the membership value of the item to obtain the membership degree of the item to each candidate classification; according to the membership degree of the item to each candidate classification, the final classification result of the item is determined through defuzzification processing, and the best sorting category of the item is determined; According to the best sorting category of items, a complete set of fuzzy control rule base is constructed, and the software algorithm of the fuzzy controller is realized by using Mamdani reasoning and center of gravity defuzzification method; the actuator is selected to realize the sorting action of items, and according to the control command signal output by the fuzzy controller and the physical parameter model of the actuator, the fuzzy PID control and fuzzy sliding mode control algorithms are used to adjust the position, speed and acceleration of the actuator in real time to optimize the sorting trajectory of items; Based on the item attribute feature vector, a fuzzy set describing different sorting categories is constructed. For the fuzzy set of each item attribute dimension, a composite membership function is constructed. The steps of achieving the continuity of the composite membership function at the boundary of adjacent sub-intervals through smooth transition technology include: Obtain multidimensional attribute data of the items to be sorted, and divide the attribute dimension into several non-intersecting sub-intervals according to the value range and data distribution pattern of each attribute dimension, each sub-interval corresponding to a fuzzy set; wherein the division of sub-intervals comprehensively considers the distribution density and discrete degree characteristics of the attribute data in this dimension; For each fuzzy set of each attribute dimension, one or more function types are selected from a preset candidate basic membership function library as the basic membership function of the subinterval according to the clustering characteristics of the data points in the corresponding subinterval; the candidate basic membership function library includes a triangular function, a trapezoidal function and a Gaussian function; By using the least square method and gradient descent parameter estimation method, the attribute data samples in the subinterval are used as the training set, and the optimal parameter set of the selected basic membership function is adaptively estimated by optimizing the objective function, so as to minimize the fitting error of the data distribution in the subinterval. The basic membership functions of all sub-intervals of each attribute dimension are combined and integrated to construct a composite membership function on the attribute dimension. The basic membership functions of adjacent sub-intervals are locally adjusted and optimized by using the smooth transition technology of interpolation fitting and spline function approximation. The multi-dimensional attribute feature vector of the item is input into the composite membership function of the corresponding attribute dimension, the membership value of the item on each fuzzy set is calculated, and the nonlinear mapping relationship between the item attribute and the fuzzy classification is constructed.

2. The method according to claim 1, characterized in that The steps of using the least squares method and gradient descent parameter estimation method, taking the attribute data samples in the subinterval as the training set, and adaptively estimating the optimal parameter set of the selected basic membership function by optimizing the objective function so as to minimize the fitting error of the data distribution in the subinterval include: For a certain attribute dimension of the item to be classified, all attribute data points in each subinterval after the attribute dimension is divided are obtained, and the attribute data points are used as a training sample set, where each sample point includes the attribute value of the point and a preset true membership value; According to the type of the selected basic membership function, an optimization objective function is constructed with the goal of minimizing the sum of squares of fitting errors. The objective function is used to measure the degree of fit of the basic membership function to the sample points in the subinterval. Initialize the parameter vector of the basic membership function. For the triangular function and trapezoidal function, the parameter vector includes the left and right boundaries and the peak point. For the Gaussian function, the parameter vector includes the mean and standard deviation. The constructed objective function is iteratively optimized using the gradient descent method. The gradient vector of the objective function with respect to each parameter is calculated through partial derivatives, and the parameters are updated along the negative gradient direction until the objective function value converges or the maximum number of iterations is reached. The parameter vector at convergence after iterative optimization is used as the optimal parameter estimate of the basic membership function in this subinterval, so as to minimize the fitting error of the attribute data distribution in this subinterval and obtain the basic membership function after parameter optimization.

3. The method according to claim 1, characterized in that The steps of combining and integrating the basic membership functions of all subintervals of each attribute dimension to construct a composite membership function on the attribute dimension, and using the smooth transition technology of interpolation fitting and spline function approximation to locally adjust and optimize the basic membership functions of adjacent subintervals include: According to the order of sub-intervals, the basic membership functions after parameter optimization on each sub-interval are sequentially spliced ​​to form a piecewise function covering the entire attribute value range; For the intersection points of adjacent sub-intervals, a smooth transition technology of interpolation fitting or spline function approximation is used to construct a transition function that meets the continuity and smoothness conditions; The smoothed piecewise function and transition function are integrated to construct a composite membership function covering the entire attribute dimension, so that it is consistent with the optimized basic membership function within the sub-interval and achieves continuous and smooth transition at the junction of adjacent sub-intervals. On the premise of maintaining continuity and smoothness, the local shape parameters of the composite membership function are fine-tuned, an additional correction function is introduced, appropriate regularization constraints are set, and numerical optimization methods are used to locally adjust and optimize the composite membership function.

4. The method according to claim 1, characterized in that: The steps of inputting the multi-dimensional attribute information of the item into the composite membership function of the corresponding attribute dimension, calculating the membership value of the item on each fuzzy set, building a fuzzy classification rule base, and using the fuzzy reasoning method to comprehensively evaluate the membership value of the item to each candidate classification include: The multi-dimensional attribute information of the items to be sorted is extracted and input into the composite membership function of the corresponding attribute dimension. The membership value of the fuzzy set corresponding to each attribute dimension of the item is calculated by the composite membership function. The membership value indicates the degree of belonging of the item to the fuzzy set corresponding to the attribute dimension. Construct a fuzzy classification rule library. The fuzzy classification rule is composed of the combination conditions of the attribute membership values ​​of the items as the antecedent and the classification results of the items as the consequent, which is used to describe the fuzzy mapping relationship between the item attributes and the classification categories. For each fuzzy classification rule, the attribute membership value of the item is substituted into the antecedent part of the fuzzy classification rule, and the trigger strength of each fuzzy classification rule is calculated by the fuzzy reasoning method. The trigger strength is calculated by the combination of the attribute membership values ​​of each item, which indicates the degree to which the fuzzy classification rule is satisfied by the item attributes. For each candidate classification category, summarize the reasoning results of all fuzzy classification rules pointing to this category, and take the maximum trigger strength of the item in all fuzzy classification rules pointing to this category as the item's membership degree to this category. The membership degree represents the belonging measure of the item to this classification category based on fuzzy reasoning. According to the degree of belonging of the items to each candidate classification category, a decision strategy is adopted to determine the final classification of the items, and the items are divided into the classification category with the highest degree of belonging.

5. The method according to claim 1, characterized in that According to the best sorting category of items, a complete set of fuzzy control rule base is constructed, and the steps of implementing the software algorithm of the fuzzy controller using Mamdani reasoning and center of gravity defuzzification method include: Input the input variable value of the current control cycle into the fuzzy controller to trigger the corresponding fuzzy control rule. According to the input variable language value combination of the antecedent part of each fuzzy control rule, calculate the membership value of each input variable corresponding to the language value, and use the minimum operation to obtain the triggering strength of each fuzzy control rule. According to the triggering strength of each fuzzy control rule, the membership function of the output variable of the consequent part of the fuzzy control rule is truncated to obtain the membership function of the output variable under each fuzzy control rule; The membership functions of all output variables that trigger the fuzzy control rules are combined, and the total membership function of the output variables is obtained by taking the largest operation; The centroid defuzzification method is used to convert the total membership function of the output variable into an accurate control quantity. The centroid defuzzification obtains the defuzzified value by calculating the center of the area under the total membership function. In the actual calculation, the total membership function is discretized into a finite number of points by numerical integration, and then the centroid is approximately calculated by weighted average. The numerical value obtained by defuzzifying the center of gravity is converted into a specific actuator control instruction, and the control instruction is sent to the actuator.

6. The method according to claim 1, characterized in that The actuator is selected to realize the sorting action of the items. According to the control command signal output by the fuzzy controller and the physical parameter model of the actuator, the fuzzy PID control and fuzzy sliding mode control algorithms are used to adjust the position, speed and acceleration of the actuator in real time. The steps of optimizing the item sorting trajectory include: Establishing a kinematic model and a dynamic model of the actuator, establishing the forward and inverse kinematic equations of the actuator by using the DH parameter method, describing the mapping relationship between the joint angle and the end position and posture, and considering the dynamic characteristics of the actuator to establish the dynamic equation of the actuator; Determine the optimization objective function and constraint conditions of the sorting trajectory, wherein the optimization objective function includes minimizing the sorting time, maximizing the average speed and minimizing the acceleration change, and the constraint conditions include the upper and lower limits of the joint angle, angular velocity and angular acceleration of the actuator and the boundary constraint of the end position; discretize the sorting trajectory, generate smooth trajectory segments between adjacent discrete nodes by interpolation method, and transform the continuous trajectory planning problem into a discrete optimization problem; A fuzzy PID control algorithm is applied, with position error and speed error as input, and the parameters of the PID controller are adjusted online through fuzzy reasoning and defuzzification calculation to generate optimized position and speed control instructions, and the optimized position and speed instructions are substituted into the kinematic model of the actuator to calculate the angle and angular velocity of each joint; a fuzzy sliding mode control algorithm is applied, with acceleration error and acceleration change rate as input, and the sliding mode surface and control law are adjusted online through fuzzy reasoning and defuzzification calculation to generate optimized acceleration control instructions, and the optimized acceleration instructions are substituted into the dynamic model of the actuator to calculate the driving torque of each joint; The sorting trajectory is iteratively optimized, and the position, velocity and acceleration instructions obtained by the fuzzy PID control algorithm and the fuzzy sliding mode control algorithm are used as the optimal values ​​of the current discrete node, substituted into the next discrete node, and the fuzzy PID control and fuzzy sliding mode control are repeatedly executed until the optimization of the entire sorting trajectory is completed.

7. A logistics classification and sorting control system based on fuzzy control, used to implement the method described in any one of claims 1 to 6, characterized in that: include: The first unit is used to obtain multi-dimensional attribute information of the items to be sorted, collect attribute data of the items through a multi-sensor array set at different positions of the sorting line, and pre-process the collected attribute data to obtain a normalized attribute feature vector of the items; The second unit is used to construct fuzzy sets describing different sorting categories based on the feature vector of item attributes, construct a composite membership function for the fuzzy set of each item attribute dimension, and realize the continuity of the composite membership function at the boundary of adjacent sub-intervals through smooth transition technology; comprehensively consider the statistical characteristics of the distribution density and discrete degree of the item attribute data points in each sub-interval, adaptively determine the basic function type and corresponding parameters of the composite membership function in each sub-interval, and establish a fuzzy mapping relationship between item attributes and sorting categories; input the multi-dimensional attribute information of the item into the composite membership function of the corresponding attribute dimension, calculate the membership value of the item on each fuzzy set, construct a fuzzy classification rule base, and use fuzzy reasoning method to comprehensively evaluate the membership value of the item to each candidate classification; according to the membership degree of the item to each candidate classification, determine the final classification result of the item through defuzzification processing, and determine the best sorting category of the item; The third unit is used to build a complete fuzzy control rule base according to the optimal sorting category of items, and use Mamdani reasoning and center of gravity defuzzification method to implement the software algorithm of the fuzzy controller; select actuators to realize the sorting action of items, and use fuzzy PID control and fuzzy sliding mode control algorithms according to the control command signal output by the fuzzy controller and the physical parameter model of the actuator to adjust the position, speed and acceleration of the actuator in real time to optimize the sorting trajectory of items.

8. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images