AGV scheduling method and system for container yard

Abstract

The invention belongs to the technical field of port scheduling, and particularly relates to a container yard AGV scheduling method and system, and the method comprises the steps: S1, collecting and processing the data of a field environment; s2, setting a target function F according to the transportation distance D, the waiting time T, the weather influence W and the storage space storage factor C; s3, using an ant colony algorithm to optimize the transportation distance D and the storage space storage factor C; optimizing the waiting time T and the weather influence W by using a neural network; and S4, selecting a scheme with the minimum comprehensive objective function F according to the transportation distance D and the storage space storage factor C provided by the ant colony algorithm and the waiting time T and the weather influence W predicted by the neural network. The method has the advantages that different key independent variables are optimized through the ant colony algorithm and the neural network respectively, and optimization of the efficiency and stability of the whole dispatching system is facilitated. According to the method and the system, a brand new algorithm is constructed for a port environment, and real-time and dynamic optimization can be realized on path and resource scheduling.

Description

Technical Field

[0001] The present application belongs to the technical field of port dispatching, and specifically relates to a method and system for dispatching AGVs in a container yard. Background Art

[0002] The traditional container yard method is to complete the loading and unloading and handling of goods through manual operation. Although this method can ensure the quality and safety of the goods, it is inefficient and prone to human factors leading to errors. Therefore, in order to improve the work efficiency and safety of the container yard, the development of automated container yard remote control systems has become an inevitable trend. Summary of the invention

[0003] Based on the above problems, the present application provides a container yard AGV scheduling method and system, and its technical solution is:

[0004] A container yard AGV scheduling method comprises the following steps:

[0005] S1. Collect and process data of the on-site environment;

[0006] S2. Set the objective function F according to the transportation distance D, waiting time T, weather impact W and warehouse storage capacity factor C;

[0007] S3. Use ant colony algorithm to optimize the transportation distance D and warehouse storage capacity factor C; use neural network to optimize waiting time T and weather impact W;

[0008] S4. According to the transportation distance D and warehouse storage capacity factor C provided by the ant colony algorithm and the waiting time T and weather impact W predicted by the neural network, select the solution with the smallest comprehensive objective function F.

[0009] Preferably, in step S2, the objective function is:

[0010] F=w 1 ·D+w 2 ·T+w 3 ·W+w 4 ·C

[0011] w 1 , W 2 、w 3 、w 4 Weight coefficient;

[0012] Constrain the transport distance D and select the best path for the container to move within the yard (including distance and optimal storage space utilization);

[0013] Constrain the waiting time T to minimize the waiting time of the container;

[0014] Weather affects W, and different weather will affect scheduling.

[0015] Storage space utilization rate C,

[0016] Preferably, in step S3, an initialization pheromone matrix τ about the transport distance D and the warehouse storage capacity factor C is constructed, and the heuristic information η is set, where η is the inverse ratio of the path distance and the warehouse balance; each container is regarded as an ant, and the path selection probability p is added to the movement rule. ij To adjust the priority of path selection.

[0017] Preferably, each ant selects a path based on τ and η, and the path selection probability p ij The calculation steps are as follows: Path congestion degree CW ij : It is set according to the real-time traffic flow of the path, and the more congested paths are given lower weights;

[0018]

[0019] TD ij : The traffic density from path i to j, which is used to indicate the number of vehicles on the path;

[0020] Path Priority PW ij : Some routes are given higher weights because they are close to some important loading and unloading areas or urgent tasks need to be handled first;

[0021] PW ij =1+k×PL ij

[0022] k is the priority adjustment coefficient, which is used to adjust the impact of the priority level on the overall weight. A larger k will make the impact of the priority level more significant;

[0023] PL ij : The priority weight of path i to j. This value is used to increase the priority of important paths in scheduling or path selection. It is determined by the port manager and three levels are defined, with values ​​of 1, 2, and 3.

[0024] Combining the above weight factors, the overall path weight is defined as:

[0025] Weight ij =CW ij ×PW ij

[0026] Combined with the weight factor, the path selection probability formula is rewritten as:

[0027]

[0028] τij (t): pheromone concentration on the path from node i to node j at time t;

[0029] η ij : The heuristic factor from node i to node j, which is the inverse of the path length;

[0030] α: adjustment parameter of pheromone importance;

[0031] β: adjustment parameter of the heuristic factor;

[0032] Weight ij : Additional weight factor for the path from node i to node j;

[0033] ∑ k∈allowed [τ ik (t)] α ·[η ik ] β Weight ik : The weighted sum of all possible paths;

[0034] Update pheromones based on path quality:

[0035]

[0036] τ ij : represents the pheromone concentration from node i to node j;

[0037] ρ: represents the pheromone volatility coefficient, with a value range of [0, 1], and is calculated in combination with the storage utilization factor of the C bin. The fuller the bin is, the greater the volatility coefficient is, which accelerates the pheromone volatility to encourage the exploration of other paths.

[0038]

[0039] ρ min : It is the minimum value of the volatility coefficient, ensuring that the pheromone will not disappear completely and remain at a certain level;

[0040] ρ max : It is the maximum value of the volatility coefficient, which is used to limit the rapid volatilization of pheromones;

[0041] is a tuning parameter used to control the sensitivity of the volatility factor to changes in bin utilization. A higher value makes the volatility factor more sensitive to high utilization.

[0042] Δτ ij : represents the pheromone increment from node i to node j, defined as:

[0043]

[0044] Q: total amount of pheromone;

[0045] L: The total length of the path traveled.

[0046] Preferably, a neural network is used to optimize the waiting time T and the weather impact W, the steps are as follows:

[0047] Step 1: Collect data: Collect container dispatch data from the historical dispatch system;

[0048] Step 2: Data cleaning: remove missing values, duplicate values ​​and outliers;

[0049] Step 3: Feature extraction:

[0050] For the T indicator: extract the dispatch time, number of containers, and yard congestion, i.e. the number of containers currently being processed;

[0051] For the W indicator: the weather conditions are numerically categorized into sunny, cloudy, rainy and stormy weather;

[0052] Step 4: Data standardization or normalization;

[0053] Step 5: Input the features into the neural network for prediction:

[0054] For T: Use the mean squared error MSE loss function:

[0055]

[0056] N is the sample size; is the predicted value, y ti is the actual value;

[0057] For W: Use the cross entropy loss function:

[0058]

[0059] Where R is the number of categories, y Ri is the one-hot encoding of the true category, is the predicted probability.

[0060] A container yard AGV dispatching system, comprising a data acquisition module, a data processing module and an output module;

[0061] Data acquisition module: LiDAR is used to detect the position of the target AGV and send the position data to the operator; data processing module: ant colony algorithm is used to optimize the transportation distance D and the warehouse storage capacity factor C; neural network is used to optimize the waiting time T and weather impact W; image analysis technology is used to analyze the position and direction of the target AGV, and the results are passed to the operator, who coordinates with the AGV to complete the corresponding task; output module: AGV scheduling information is displayed on the LED display.

[0062] Preferably, it includes automatically identifying and tracking the location and status of each AGV, and providing corresponding data analysis and report generation services.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] 1. Improve the neural network algorithm to make it more suitable for the port environment; improve the ant colony algorithm, introduce variables and adjust the formula to make it more suitable for the port environment.

[0065] 2. The system can also automatically identify and track the location and status of each container, and provide corresponding data analysis and report generation services, thereby improving work efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is the flow chart of this application.

[0067] Figure 2 Schematic diagram of an automated container yard.

[0068] Figure 3 It is the hardware structure of the automated container yard system. DETAILED DESCRIPTION

[0069] The technical solution of the present application is described in detail below through specific embodiments and drawings. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application, and the specific technical features may be combined with each other.

[0070] A container yard AGV scheduling method comprises the following steps:

[0071] S1. Collect and process data of the on-site environment;

[0072] S2. Set the objective function F according to the transportation distance D, waiting time T, weather impact W and warehouse storage capacity factor C;

[0073] The objective function is:

[0074] F=w 1 ·D+w 2 ·T+w3 ·W+w 4 ·C

[0075] w 1 、w 2 、w 3 、w 4 Weight coefficient;

[0076] Restrict the transport distance D and reduce the distance that containers move in the yard;

[0077] Constrain the waiting time T to minimize the waiting time of the container;

[0078] Weather affects W, and different weather will affect scheduling.

[0079] Storage space utilization rate C,

[0080] S3. Use ant colony algorithm to optimize transportation distance D and warehouse storage capacity factor C; use neural network to optimize waiting time T and weather impact W ；

[0081] In step S3, an initial pheromone matrix τ is constructed about the transport distance D and the warehouse storage capacity factor C, and the heuristic information η is set, where η is the inverse ratio of the path distance and the warehouse balance; each container is regarded as an ant, and the path selection probability p is added to the movement rule. ij To adjust the priority of path selection.

[0082] Each ant selects a path based on τ and η, and the path selection probability p ij The calculation steps are as follows:

[0083] Path congestion level CW ij : It is set according to the real-time traffic flow of the path, and the more congested paths are given lower weights;

[0084]

[0085] TD ij : The traffic density from path i to j, used to indicate the number of vehicles on the path.

[0086] Path Priority PW ij : Some routes are given higher weights because they are close to some important loading and unloading areas or urgent tasks need to be handled first;

[0087] PW ij =1+k×PL ij

[0088] k is the priority adjustment coefficient. This coefficient is used to adjust the impact of the priority level on the overall weight. A larger k will make the impact of the priority level more significant.

[0089] PL ij : The priority weight of path i to j. This value is used to increase the priority of important paths in scheduling or path selection. It is determined by the port manager and three levels are defined, with values ​​of 1, 2, and 3.

[0090] Combining the above weight factors, the overall path weight is defined as:

[0091] Weight ij =CW ij ×PW ij

[0092] Combined with the weight factor, the path selection probability formula is rewritten as:

[0093]

[0094] τ ij (t): pheromone concentration on the path from node i to node j at time t;

[0095] η ij : The heuristic factor from node i to node j, which is the inverse of the path length;

[0096] α: adjustment parameter of pheromone importance;

[0097] β: adjustment parameter of the heuristic factor;

[0098] Weight ij : Additional weight factor for the path from node i to node j

[0099] ∑ k∈allowed [τ ik (t)] α ·[η ik ] β Weight ik : The weighted sum of all possible paths.

[0100] Update pheromones based on path quality:

[0101]

[0102] τ ij : represents the pheromone concentration from node i to node j.

[0103] ρ: represents the pheromone volatility coefficient, with a value range of [0, 1]. It is calculated here in combination with the storage utilization factor of the C bin. The fuller the bin storage, the greater the volatility coefficient, which accelerates the pheromone volatility to encourage the exploration of other paths.

[0104] first,

[0105]

[0106] Then,

[0107]

[0108] ρ min : It is the minimum value of the volatility coefficient, which ensures that the pheromone will not disappear completely and remain at a certain level.

[0109] ρ max : It is the maximum value of the volatility coefficient, which is used to limit the rapid volatilization of pheromones.

[0110] is a tuning parameter that controls the sensitivity of the volatility factor to changes in bin utilization. Higher values ​​make the volatility factor more sensitive to high utilization.

[0111] Δτ ij : represents the pheromone increment from node i to node j, defined as:

[0112]

[0113] Q: total amount of pheromone;

[0114] L: The total length of the path traveled.

[0115] Use neural network to optimize waiting time T and weather impact W. The steps are as follows:

[0116] Step 1: Collect data: Collect container dispatch data from the historical dispatch system;

[0117] Step 2: Data cleaning: remove missing values, duplicate values ​​and outliers;

[0118] Step 3: Feature extraction:

[0119] For T indicators: extract the time to lift a single container, the total number of containers, and the number of containers currently being processed;

[0120] For the W indicator: digitize weather conditions and extract wind speed, rainfall, and ultraviolet intensity.

[0121] Step 4: Data standardization or normalization;

[0122] Step 5: Input the features into the neural network for prediction, and use different loss functions for T and W. In the iterative prediction, minimize the value of the loss function to achieve accurate prediction.

[0123] For T: Use the mean squared error MSE loss function:

[0124]

[0125] N is the sample size; is the predicted value, y ti is the actual value;

[0126] For W: Use the cross entropy loss function:

[0127]

[0128] Where R is the number of categories, y Ri is the one-hot encoding of the true category, is the predicted probability.

[0129] S4. According to the transportation distance D and warehouse storage capacity factor C provided by the ant colony algorithm and the waiting time T and weather impact W predicted by the neural network, select the solution with the smallest comprehensive objective function F.

[0130] A container yard AGV dispatching system includes a data acquisition module, a data processing module and an output module; in the data acquisition module, a laser radar is used to detect the position of a target AGV and send the position data to an operator; in the data processing module, an ant colony algorithm is used to optimize the transportation distance D and the storage capacity factor C of the warehouse; a neural network is used to optimize the waiting time T and the weather influence W; image analysis technology is used to analyze the position and direction of the target AGV, and the result is transmitted to the operator, and the operator coordinates with the AGV to complete the corresponding task; in the output module, the AGV dispatching information is displayed on an LED display screen;

[0131] Automatically identify and track the location and status of each AGV, and provide corresponding data analysis and report generation services.

[0132] In this way, the ant colony algorithm and the neural network optimize different key independent variables respectively, which helps to optimize the efficiency and stability of the entire scheduling system. Both have constructed a new algorithm for the port environment, which can not only achieve real-time and dynamic optimization in path and resource scheduling, but also improve the accuracy and efficiency of scheduling by learning historical patterns. This collaborative strategy provides an intelligent and adaptable optimization framework for container scheduling, which can significantly improve operational efficiency and cost-effectiveness.

[0133] 1. System hardware composition:

[0134] The hardware of the automated container yard remote control system mainly consists of a control unit module, a sensor module, a mobile terminal device, a power supply module, a communication module, a motor drive module, etc. Among them, the control unit module is the core component of the entire system, which is responsible for receiving the user's operation instructions and performing corresponding processing and execution. The module uses the ARMCortex-A9 processor as the core, with high-performance computing capabilities and powerful data processing functions. At the same time, it is also equipped with a high-speed storage chip and a high-speed communication interface, which can quickly respond to user requests and realize data collection and transmission of the on-site environment. The sensor module is an important component for collecting environmental parameters, including temperature sensors, humidity sensors, and air pressure sensors. These sensors are connected to the server-side device through the RS-485 serial port. The mobile terminal device is used to realize real-time monitoring and management of the yard. The mobile terminal device can use Wi-Fi or Bluetooth for wireless communication and supports GPS positioning function. The power supply module provides a stable power supply for the entire system, and the UPS power supply method is used to ensure the stability of the system. The communication module is mainly used to transmit data and instructions, upload data to the cloud platform through the GPRS / CDMA network, and can also receive data and instructions from the cloud platform. Finally, the motor drive module converts the signal into actual action torque. It can control the speed and direction of the motor through PWM signals to complete the mechanical operation task.

[0135] 2. System software structure:

[0136] A scheme of remote control system for automated container yard based on Internet of Things is proposed. The system mainly consists of three parts: sensor acquisition module, data transmission module and control center module. Among them, the sensor acquisition module is responsible for collecting and processing data of the on-site environment; the data transmission module sends these data to the control center module through the wireless network, and the control center module is one of the core components of the whole system. It can receive data from the sensor acquisition module and convert it into control signals to realize remote control of vehicles in the yard. In order to ensure the stability and reliability of the system, full consideration is given to the design of the system software. Specifically, the system software adopts a distributed architecture mode, that is, the various functional modules of the system are deployed on different servers respectively to improve the scalability and flexibility of the system. At the same time, some advanced programming languages ​​and development tools are used to build the software framework of the system, such as Java, Spring, MyBatis, etc. In addition, some security protection measures are added to the system, such as identity authentication mechanism, encryption communication protocol, etc., to ensure the security and confidentiality of the system.

[0137] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A container yard AGV scheduling method, characterized in that: The following steps are involved: S1. Collect and process data of the on-site environment; S2. Set the objective function F according to the transportation distance D, waiting time T, weather impact W and warehouse storage capacity factor C; S3. Use ant colony algorithm to optimize the transportation distance D and warehouse storage capacity factor C; use neural network to optimize waiting time T and weather impact W; S4. According to the transportation distance D and warehouse storage capacity factor C provided by the ant colony algorithm and the waiting time T and weather impact W predicted by the neural network, select the solution with the smallest comprehensive objective function F.

2. The container yard AGV dispatching method according to claim 1, characterized in that: In step S2, the objective function is: F=w1·D+w2·T+w3·W+w4·C w1, w2, w3, w4 weight coefficients; Constrain the transport distance D and select the best path for the container to move within the yard; Constrain the waiting time T to minimize the waiting time of the container; Weather affects W, and different weather conditions can affect scheduling; Storage space utilization rate C, 3. The container yard AGV dispatching method according to claim 1, characterized in that: In step S3, an initial pheromone matrix τ is constructed about the transport distance D and the warehouse storage capacity factor C, and the heuristic information η is set, where η is the inverse ratio of the path distance and the warehouse balance; each container is regarded as an ant, and the path selection probability p is added to the movement rule. ij To adjust the priority of path selection.

4. The container yard AGV dispatching method according to claim 3, characterized in that: Each ant selects a path based on τ and η, and the path selection probability p ij The calculation steps are as follows: Path congestion level CW ij : It is set according to the real-time traffic flow of the path, and the more congested paths are given lower weights; TD ij : The traffic density from path i to j, which is used to indicate the number of vehicles on the path; Path Priority PW ij : Some routes are given higher weights because they are close to some important loading and unloading areas or urgent tasks need to be handled first; PW ij =1+k×PL ij k is the priority adjustment coefficient; PL ij : The priority weight of path i to j. This value is used to increase the priority of important paths in scheduling or path selection. It is determined by the port manager and three levels are defined, with values ​​of 1, 2, and 3. Combining the above weight factors, the overall path weight is defined as: Weight ij =CW ij ×PW ij Combined with the weight factor, the path selection probability formula is rewritten as: τ ij (t): pheromone concentration on the path from node i to node j at time t; η ij : The heuristic factor from node i to node j, which is the inverse of the path length; α: adjustment parameter of pheromone importance; β: adjustment parameter of the heuristic factor; Weight ij : Additional weight factor for the path from node i to node j; ∑ k∈allowed [τ ik (t)] α ·[η ik ] β Weight ik : The weighted sum of all possible paths; Update pheromones based on path quality: τ ij : represents the pheromone concentration from node i to node j; ρ: represents the pheromone volatility coefficient, with a value range of [0,1], and is calculated in combination with the storage utilization factor of the C bin. The fuller the bin is, the greater the volatility coefficient is, which accelerates the pheromone volatility to encourage the exploration of other paths. ρ min : It is the minimum value of the volatility coefficient, ensuring that the pheromone will not disappear completely and remain at a certain level; ρ max : It is the maximum value of the volatility coefficient, which is used to limit the rapid volatilization of pheromones; is a tuning parameter used to control the sensitivity of the volatility factor to changes in bin utilization. A higher value makes the volatility factor more sensitive to high utilization. Δτ ij : represents the pheromone increment from node i to node j, defined as: Q: total amount of pheromone; L: The total length of the path traveled.

5. The container yard AGV dispatching method according to claim 1, characterized in that: Use neural network to optimize waiting time T and weather impact W. The steps are as follows: Step 1: Collect data: Collect container dispatch data from the historical dispatch system; Step 2: Data cleaning: remove missing values, duplicate values ​​and outliers; Step 3: Feature extraction: For the T indicator: extract the dispatch time, number of containers, and yard congestion, i.e. the number of containers currently being processed; For the W indicator: the weather conditions are numerically categorized into sunny, cloudy, rainy and stormy weather; Step 4: Data standardization or normalization; Step 5: Input the features into the neural network for prediction: For T: Use the mean squared error MSE loss function: N is the sample size; is the predicted value, y ti is the actual value; For W: Use the cross entropy loss function: Where R is the number of categories, y Ri is the one-hot encoding of the true category, is the predicted probability.

6. A container yard AGV dispatching system, using any method according to claims 1-5, characterized in that: It includes a data acquisition module, a data processing module and an output module; Data acquisition module: LiDAR is used to detect the position of the target AGV and send position data to the operator; Data processing module: Use ant colony algorithm to optimize the transportation distance D and warehouse storage capacity factor C; use neural network to optimize waiting time T and weather impact W; use image analysis technology to analyze the position and direction of the target AGV, and pass the results to the operator, who coordinates with the AGV to complete the corresponding task; Output module: AGV scheduling information is displayed on the LED display.

7. The container yard AGV dispatching system according to claim 6, characterized in that: It includes automatic identification and tracking of the location and status of each AGV, and provides corresponding data analysis and report generation services.

Images