An AGV Scheduling Method Based on a Time Estimation Model

Abstract

The present invention provides an AGV scheduling method based on a time estimation model. The method includes the following steps: a static estimation step of estimating a static estimated value of the bridge crane work efficiency and a static estimated value of the rail crane work efficiency, and estimating a static estimated time of the container at each work stage; a dynamic estimation step of estimating a dynamic estimated value of the bridge crane work efficiency and a dynamic estimated value of the rail crane work efficiency, and estimating a dynamic estimated time of the container at each work stage; and a final estimated time of the container at each work stage and a total estimated time of the total work are calculated based on the static estimated time and the dynamic estimated time of the container at each work stage, and an AGV work scheduling command is generated based on the final estimated time of the container at each work stage and the total estimated time of the total work. The AGV scheduling method based on the time estimation model of the present invention combines the estimated time to schedule the AGV more reasonably, shorten the waiting time of the AGV, shorten the unloading travel time and distance of the AGV, and improve the container transport efficiency of the AGV.

Description

[Technical field]

[0001] The present invention belongs to the field of transportation technology of automated terminals, and particularly relates to an AGV scheduling method based on a time estimation model. [Background technology]

[0002] The container terminal front work zone refers to the area between the yard front side line and the quay front line, its functions are quay bridge cargo ship work and container entry and exit yard work service, artificial quay and semi-automated terminal usually use the inner set card of artificial operation. In recent years, with the increase of labor costs, more and more terminals have converted the horizontal transportation equipment from the inner set card to AGV (Automated Guided Vehicle), and the use of AGV has greatly improved production efficiency and saved costs.

[0003] On the sea side of the yard of a fully automated container terminal, a bridge enables interaction between cargo ships and automated guide vehicles, a rail crane enables interaction between the automated guide vehicles and the yard, and AGVs, L-AGVs, Auto Shuttles, etc. handle automated operations between the quay and the yard.

[0004] At present, the rational scheduling of AGVs, the stability of the system scheduling, and the maximization of the utilization rate of AGVs are the keys to improving the production efficiency of the terminal. The working status of AGVs is also an important part of the work of the wharf. Real-time monitoring of the working status of AGVs can enable workers to find problems in the production process earlier, improve and perfect the scheduling algorithm of AGVs, and improve the work efficiency of the entire automated terminal. Summary of the Invention [Problem to be solved by the invention]

[0005] The present invention solves the technical problem of low efficiency in transporting containers by AGVs in the prior art, and proposes an AGV scheduling method based on a time estimation model that can solve the above problem. [Means for solving the problem]

[0006] In order to achieve the above technical effects, the present invention adopts the following technical solutions:

[0007] The AGV scheduling method based on the time estimation model includes the following steps: a static estimation step of estimating a static estimated value of bridge crane work efficiency and a static estimated value of rail crane work efficiency, and estimating a static estimated time of the container at each work stage, including a static estimated time of an interaction stage with the bridge crane work, a static estimated time of an interaction stage with the rail crane work, a static estimated time of an AGV travel stage, a static estimated time of a bridge crane work stage, and a static estimated time of a rail crane work stage; a dynamic estimation step of estimating a dynamic estimated value of a bridge crane work efficiency and a dynamic estimated value of a rail crane work efficiency, and estimating a dynamic estimated time of the container at each work stage, including a dynamic estimated time of an interaction stage with the bridge crane work, a dynamic estimated time of an interaction stage with the rail crane work, a dynamic estimated time of an AGV traveling stage, a dynamic estimated time of a bridge crane work stage, and a dynamic estimated time of a rail crane work stage; Calculate a final estimated time of the container at each work stage and a total estimated work time based on the static estimated time and the dynamic estimated time of each work stage, and the total estimated work time is the sum of the final estimated times of each work stage; Generate AGV work scheduling instructions based on the estimated final container time at each work stage and the estimated total work time.

[0008] Furthermore, in the static estimation step, the static estimation values ​​for estimating the bridge crane work efficiency and the rail crane work efficiency include static estimation values ​​for estimating the bridge crane work efficiency and the rail crane work efficiency by acquiring historical data of work efficiency, The method for estimating the static estimated time of the AGV travel phase is as follows: Obtain the historical time of the AGV travel phase, and estimate the static estimated time of the AGV travel phase; Estimating a static estimated time for a bridge crane work phase and a static estimated time for a rail crane work phase based on the static estimated value of the bridge crane work efficiency and the static estimated value of the rail crane work efficiency; Based on the static estimated value of the bridge crane work efficiency, the static estimated value of the rail crane work efficiency, and the static estimated time of the AGV traveling phase, the AGV estimates the static estimated time of the interaction phase with the bridge crane work and the static estimated time of the interaction phase with the rail crane work, respectively.

[0009] Furthermore, a final estimated time and a total estimated work time for each container in each work stage are calculated based on the static estimated time and the dynamic estimated time for each work stage using a weighted average method.

[0010] Furthermore, the dynamic estimation step further includes a step of training (learning) the operation efficiency estimation model, which includes obtaining a history time of the bridge crane operation stage of each container and a position of the corresponding container; A work efficiency estimation model is trained (learned) based on the history time and the position, and the work efficiency estimation model is a multi-valued linear regression equation; The work efficiency estimation model is

number

[0011] Furthermore, the method of specifying w and b in the work efficiency estimation model is as follows: array Construct TIFF0007667862000003.tif8161, Time Loss Calculate TIFF0007667862000004.tif11162, where y i denotes the actual work time of the container at position i in the current work step, Determine w and b so that J is minimized.

[0012] Furthermore, a dynamic estimated time for a bridge crane work phase and a dynamic estimated time for a rail crane work phase are generated based on the dynamic estimated values ​​of bridge crane work efficiency and rail crane work efficiency output by the work efficiency estimation model and an input scheduling system of the number of containers that need to be transported.

[0013] Additionally, the dynamic estimation step further includes training an AGV travel time estimation model.

[0014] Further, the AGV travel time estimation model is trained based on the history data and the position, and the AGV travel time estimation model is a polynomial regression equation, and the polynomial regression equation is

number

[0015] Furthermore, before generating the AGV work scheduling command, a work priority of the container is determined, and an AGV work scheduling command is generated based on the work priority; The setting of the work priority is based on a time response ratio, TIFF0007667862000007.tif14161, The greater the time response ratio, the lower the task priority.

[0016] Furthermore, the algorithms of mixed integer programming, combinatorial optimization method and collaborative filtering are used to generate AGV work schedule commands. Effect of the Invention

[0017] Compared with the prior art, the advantages and positive effects of the invention are that the AGV scheduling method based on the time estimation model of the present invention uses a combined method of static estimation and dynamic estimation to estimate the time when a container is used at each work stage, and combined with the estimated time, makes the scheduling of AGVs more reasonable, reduces the AGV waiting time, shortens the AGV unloading operation time and distance, and improves the AGV transportation efficiency. [Brief description of the drawings]

[0018] In order to more clearly describe the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings necessary for the embodiments. Obviously, the drawings in the following description are some embodiments of the present invention, and those skilled in the art can obtain other drawings from these drawings without creative efforts.

[0019] [Figure 1] 1 is a flowchart showing an embodiment of an AGV scheduling method based on a time estimation model proposed by the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0020] Preferred Embodiments of the Invention In order to make the objectives, technical solutions and advantages of the present invention clearer, specific embodiments of the present invention will now be described in more detail in conjunction with the accompanying drawings.

[0021] In describing the present invention, it should be noted that terms indicating directions or positional relationships, such as "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., are based on the directions or positional relationships shown in the accompanying drawings. These are for illustrative purposes only and do not indicate or suggest that the device or element in question must have a particular orientation, be configured or operate in a particular way, and therefore cannot be construed as limiting the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance.

[0022] Example 1 The process of loading and unloading cargo at a container terminal is to first transfer the container from the cargo ship to the AGV by a bridge crane, then the AGV transports the container to the yard, and the rail crane stacks the container at a designated position in the yard. However, the number of AGVs is limited, and if the scheduling is not feasible, the waiting time of the AGVs will be long, resulting in poor transportation efficiency. In order to solve the above problem, this embodiment proposes an AGV scheduling method based on a time estimation model, which includes the following steps as shown in Figure 1.

[0023] a static estimation step of estimating a static estimated value of bridge crane work efficiency and a static estimated value of rail crane work efficiency, and estimating a static estimated time of the container at each work stage, including a static estimated time of an interaction stage with the bridge crane work, a static estimated time of an interaction stage with the rail crane work, a static estimated time of an AGV travel stage, a static estimated time of a bridge crane work stage, and a static estimated time of a rail crane work stage; a dynamic estimation step of estimating a static estimated value of bridge crane work efficiency and a static estimated value of rail crane work efficiency, and estimating a dynamic estimated time of the container at each work stage, including a dynamic estimated time of an interaction stage with the bridge crane work, a dynamic estimated time of an interaction stage with the rail crane work, a dynamic estimated time of an AGV traveling stage, a dynamic estimated time of a bridge crane work stage, and a dynamic estimated time of a rail crane work stage; Calculate a final estimated time of the container at each work stage and a total estimated work time according to the static estimated time and the dynamic estimated time of each work stage, and the total estimated work time is the sum of the final estimated times of each work stage; Generate AGV work scheduling commands based on the estimated final container time at each work stage and the estimated total work time.

[0024] The AGV scheduling method based on the time estimation model uses a combined method of static estimation and dynamic estimation to estimate the time that a container will be used at each work stage, and schedules AGVs more reasonably according to the estimated time, thereby reducing AGV waiting time, shortening AGV unloading operation time and distance, and improving the container transportation efficiency of AGVs.

[0025] The time estimation for each work step using a method combining static setting and dynamic adjustment can not only be adjusted manually according to the actual production work situation, but also can use the dynamic adjustment method to improve and perfect the forward / backward sequence of AGV scheduling, thereby improving the work efficiency of the entire terminal.

[0026] In a preferred embodiment, in the static estimation step, the static estimation values ​​for estimating the bridge crane working efficiency and the rail crane working efficiency respectively include the following steps:

[0027] Obtain the historical data of work efficiency, and estimate the static estimation value of bridge crane work efficiency and the static estimation value of rail crane work efficiency respectively. The estimation of bridge crane work efficiency and rail crane work efficiency is mainly based on two variables, namely, 1, the current actual work situation, and 2, the historical data. The operator combines the two variables to perform a static estimation of work efficiency. The estimation can be performed using the weighted average method.

[0028] The method for estimating the static estimated time of the AGV travel phase is as follows: The historical time of the AGV travel phase is obtained, and the static estimated time of the AGV travel phase is estimated.

[0029] Based on the static estimated time of the bridge crane work phase and the static estimated time of the rail crane work phase, the static estimated value of the bridge crane work phase and the static estimated value of the rail crane work efficiency are estimated. The AGV travel time is estimated based on two variables, namely, 1, the current actual work situation, and 2, a fixed time matrix, which is based on historical data, and the AGV travel time is estimated by combining the two variables.

[0030] Based on the static estimated value of the bridge crane work efficiency, the static estimated value of the rail crane work efficiency, and the static estimated time of the AGV's traveling phase, the static estimated time of the AGV's interaction phase with the bridge crane work and the static estimated time of the interaction phase with the rail crane work are estimated, respectively.

[0031] In this embodiment, it is preferable to calculate the final estimated time and the total estimated work time of each container at each work stage by a weighted average method based on the static estimated time and the dynamic estimated time of each work stage.

[0032] In this step, big data processing related algorithms are used to dynamically estimate bridge crane work efficiency and rail crane work efficiency, estimated AGV running time, estimated AGV interaction time with bridge crane and estimated AGV interaction time with rail crane.

[0033] Specifically, the dynamic estimation step further includes training a work efficiency estimation model, the training step including obtaining a history time of the bridge crane work stage of each container and a position of the corresponding container; The work efficiency estimation model is trained based on the history time and the position, and the work efficiency estimation model is a multi-valued linear regression equation, and the work efficiency estimation model is

number

[0034] The dynamic estimation of bridge crane work efficiency and the dynamic estimation of rail crane work efficiency use a multi-valued linear regression method to calculate the historical time of each container when performing bridge crane work from historical data, which is used as a dependent variable, and combine the historical position data corresponding to each container as an argument, and when the efficiency of the bridge crane remote operator is relatively stable, the argument and the dependent variable have a linear relationship. After using multiple linear regression models to calculate parameters from historical data samples, the time required for the estimated work can be estimated based on the current container position information.

[0035] The method of specifying w and b in the work efficiency estimation model is as follows: array Construct TIFF0007667862000010.tif8161 Time Loss Calculate TIFF0007667862000011.tif11162, where y i denotes the actual work time of the container at position i in the current work step, Determine w and b so as to minimize J.

[0036] A dynamic estimated time for the bridge crane work phase and a dynamic estimated time for the rail crane work phase are generated based on the dynamic estimated value of the bridge crane work efficiency output from the work efficiency estimation model, the dynamic estimated value of the rail crane work efficiency, and the input scheduling system of the number of containers that need to be transported.

[0037] The dynamic estimation step further includes training an AGV travel time estimation model.

[0038] The AGV travel time estimation model is trained based on the history data and the position, and the AGV travel time estimation model is a polynomial regression equation, and the polynomial regression equation is:

number

[0039] The AGV travel time estimation is a method that employs polynomial regression, and information on whether the AGV needs to make a U-turn or not is already given, given the determined AGV travel path between two points.

[0040] (w j , b) (j = 1, , n) are the estimated parameters. x in the above formula j Instead of z j Using z=(z1,z2,...,z n ) is equivalent to finding a linear function h(z), and the loss is calculated by fitting the data using the linear function h(z). The training goal is to find a set of parameters that minimizes the loss function. The answer is to find TIFF0007667862000015.tif10162.

[0041] It is also possible to remove unimportant nonlinear relationships by adding an L1 regularization term.

number

[0042] In this algorithm, the time data in the history samples is directly calculated as Y, the start point, end point, and U-turn information (turning point information) are X, and the parameters are updated in the direction that minimizes the loss function.

[0043] The interaction time of the AGV with the bridge crane is estimated based on the bridge crane work efficiency and the AGV running time estimation, and data information such as historical interaction time and interactive area distance are simultaneously combined, and the time estimation is performed using a Gaussian process regression model.

[0044] The interaction time estimation of the AGV with the bridge crane is based on the ASC work efficiency in S21 and the AGV running time estimation in S22, and simultaneously combines data information such as historical interaction time and interactive area distance, and performs time estimation using a Gaussian process regression model.

[0045] After testing with the training set and the test set, the historical data is introduced into the regression model to obtain a Gaussian regression curve to estimate the interaction time of the AGV with the rail crane.

[0046] After testing using the training set and the test set, the historical data is introduced into the regression model and a Gaussian regression curve is obtained to estimate the interaction time of the AGV with the bridge crane.

[0047] Before generating the AGV work scheduling command, determine a work priority of the container, and generate an AGV work scheduling command based on the work priority; The setting of the work priority is based on a time response ratio, TIFF0007667862000017.tif13161, The greater the time response ratio, the lower the task priority.

[0048] Mixed integer programming, combinatorial optimization method and collaborative filtering algorithm are used to generate AGV work scheduling instructions.

[0049] This dynamic task priority adjustment method is compatible with taking into account short tasks while overcoming task starvation, and can obtain timely processing even for long tasks.

[0050] The above embodiments are merely for illustrating the technical solutions of the present invention, and are not limited thereto. The present invention has been described in detail with reference to the above embodiments. Those skilled in the art may modify the technical solutions described in the above embodiments or replace some of the technical features with equivalents, but such modifications and replacements do not deviate from the essence of the corresponding technical solutions and the spirit and scope of the technical solutions that the present invention is intended to protect.

Claims

1. A method for scheduling AGVs based on a time estimation model, comprising the steps of: a step of estimating a static estimated value of bridge crane work efficiency and a static estimated value of rail crane work efficiency, and estimating a static estimated time of the container at each work stage, the static estimated time being estimated for each of the work stages, including a static estimated time of a stage where an AGV interacts with the work of the bridge crane, a static estimated time of a stage where an AGV interacts with the work of the rail crane, a static estimated time of an AGV traveling stage, a static estimated time of a work stage of the bridge crane, and a static estimated time of a work stage of the rail crane; a step of estimating a dynamic estimated value of the bridge crane work efficiency and a dynamic estimated value of the rail crane work efficiency, and estimating a dynamic estimated time of the container at each work stage, the dynamic estimated time being a dynamic estimation step of estimating a dynamic estimated time of the container at each work stage, the dynamic estimated time being a dynamic estimation time of a stage where an AGV interacts with the work of the bridge crane, a dynamic estimated time of a stage where an AGV interacts with the work of the rail crane, a dynamic estimated time of an AGV traveling stage, a dynamic estimated time of a work stage of the bridge crane, and a dynamic estimated time of a work stage of the rail crane; Calculate a final estimated time during which the container is transported at each work stage and an estimated total work time using a weighted average method based on the static estimated time of the container at each work stage and the dynamic estimated time of the container, and the estimated total work time is the sum of the final estimated times of each work stage; generating an AGV work scheduling command based on the final estimated time for the container at each work stage and the total work estimated time; A method for scheduling AGVs based on a time estimation model, comprising:

2. In the static estimation step, historical data of work efficiency is acquired to obtain a static estimated value of the bridge crane work efficiency and a static estimated value of the rail crane work efficiency, and the bridge crane work efficiency and the rail crane work efficiency are estimated using a weighted average method, respectively; The method for estimating the static estimation time of the AGV traveling phase includes the steps of: Obtaining a historical time of an AGV travel phase and estimating a static estimated time of the AGV travel phase; estimating a static estimated time for the bridge crane work phase and a static estimated time for the rail crane work phase based on the static estimated value of the bridge crane work efficiency and the static estimated value of the rail crane work efficiency; Based on the static estimated value of the bridge crane work efficiency, the static estimated value of the rail crane work efficiency, and the static estimated time of the AGV traveling phase, a static estimated time of a phase in which the AGV interacts with the work of the bridge crane and a static estimated time of a phase in which the AGV interacts with the work of the rail crane are estimated.

2. The method for AGV scheduling based on a time estimation model according to claim 1.

3. 3. The AGV scheduling method based on a time estimation model according to claim 2, wherein the final estimated time and the total work estimated time of each container in each work stage are calculated based on the static estimated time and the dynamic estimated time of each work stage using a weighted average method.

4. The dynamic estimation step further includes a step of training an operation efficiency estimation model, the step including obtaining a history time of the bridge crane operation stage of each container and a position of the corresponding container; The work efficiency estimation model is trained based on the historical time and the position, and the work efficiency estimation model is a multi-valued linear regression equation, and the work efficiency estimation model is [0010] and Where: 【number】 is the history time of each container at the current work stage, k is the number of containers and is a positive integer, f(x) is the dynamic estimated time output by the work efficiency estimation model, w is the parameter matrix of x, and b is a constant parameter; 2. The method for AGV scheduling based on a time estimation model according to claim 1, further comprising estimating the dynamic estimated value of the bridge crane work efficiency based on w and b.

5. The method of specifying w and b in the work efficiency estimation model is as follows: array Build Time Loss Calculate Here, y ｉ denotes the actual work time of the container at the position i in the current work step, 5. The AGV scheduling method based on a time estimation model according to claim 4, wherein w and b are determined so as to minimize J.

6. 6. The AGV scheduling method based on a time estimation model according to claim 5, further comprising generating a dynamic estimated time for the bridge crane work stage based on a dynamic estimated value of the bridge crane work efficiency output by the work efficiency estimation model and an input scheduling system of a number of containers requiring transportation.

7. 6. The method of claim 5, wherein the dynamic estimation step further comprises training an AGV travel time estimation model.

8. The AGV travel time estimation model is trained based on historical data and the location, the AGV travel time estimation model is a polynomial regression equation, the polynomial regression equation being: [0025] and Where: 【number】 is the travel time history data between the bridge crane and the rail crane of the AGV, k is the number of containers and is a positive integer, g(x) is the dynamic estimate of the travel phase output by the AGV travel time estimation model, w ｊ xxx ｊ 8. The AGV scheduling method based on a time estimation model according to claim 7, wherein b1 is a parameter and b2 is a constant parameter.

9. Before generating the AGV work scheduling command, a work priority of the container is determined, and the AGV work scheduling command is generated based on the work priority; The setting of the work priority is based on a time response ratio; and 7. The AGV scheduling method based on a time estimation model according to claim 1, wherein the larger the time response ratio is, the lower the task priority is.

10. 10. The AGV scheduling method based on a time estimation model according to claim 9, further comprising generating the AGV work scheduling command using one selected from the group consisting of a mixed integer programming, a combinatorial optimization method, and a collaborative filtering algorithm.

Images