Annular RGV automatic control system and method

Abstract

The invention provides an annular RGV automatic control system and method, and relates to the field of intelligent control, and the method comprises the steps: firstly receiving task information sent by an RGV scheduling system, distributing the task information to RGVs, and then switching the operation modes of the RGVs according to the task information and the real-time states of other trolleys on an annular track. And finally, guiding the RGV to execute an action corresponding to the operation mode so as to complete the operation task. In this way, effective scheduling and control of the RGVs can be achieved, and then the logistics transportation efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent control, and more specifically, to a ring RGV automatic control system and method. Background Art

[0002] Circular RGV (Circular Guided Vehicle) is an automated material handling device that runs along a fixed circular track and is used to transport materials from one point to another. It is usually used inside factories or warehouses to achieve efficient, accurate and automated material transportation. However, when multiple RGVs turn on the same track, they may face problems such as low efficiency, low control accuracy and downtime caused by adjacent vehicles waiting for each other.

[0003] In this regard, patent CN117707173A proposes a circular track RGV motion control method. Specifically, the RGV scheduling system receives task control instructions and assigns tasks to the RGV vehicle. The task control unit changes the control mode of the RGV vehicle through switching instructions according to the assigned tasks and the operating conditions of other RGVs on the circular track to complete the corresponding actions, thereby improving the operating efficiency of the entire system.

[0004] The traditional control mode (operating mode) judgment method described in the patent mainly relies on rule-based logical judgment and simple vehicle spacing and status information processing. However, this method relies on fixed rules and thresholds and cannot flexibly respond to complex and changing environments. For example, under different workloads or operating conditions, the same vehicle spacing threshold may no longer apply, resulting in the system being unable to optimally adjust the behavior of the RGV car. In addition, fixed threshold settings (such as specific values ​​of vehicle spacing) may lead to misjudgments or untimely responses in dynamic environments. For example, when the vehicle spacing threshold fails to accurately reflect the actual operating conditions, it may erroneously trigger a deceleration or stop command, affecting efficiency; or when an immediate response is required, the strategy cannot be adjusted quickly due to reliance on predetermined rules, resulting in potential safety hazards or traffic jams, reducing the flexibility and reliability of the system.

[0005] Therefore, an optimized ring RGV control scheme is desired. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a ring RGV automatic control system and method.

[0007] According to one aspect of the present application, a circular RGV automatic control system is provided, which includes: a task allocation module for receiving task information issued by an RGV scheduling system and allocating the task information to an RGV trolley; a mode switching module for switching the operation mode of the RGV trolley according to the task information and in combination with the real-time operation status of other RGV trolleys on the circular track; a trolley execution module for instructing the RGV trolley to execute actions corresponding to the operation mode to complete the operation task; wherein the mode switching module includes: a trolley operation data acquisition unit for acquiring a timing queue of the operation status data of a target RGV trolley object, and acquiring the real-time operation status data of a preceding vehicle of the target RGV trolley object on the circular track to obtain a timing queue of the operation status data of the preceding vehicle of the RGV; an embedding unit for performing an operation on the timing queue of the operation status data of the target RGV trolley object and the operation status data of the preceding vehicle of the RGV The temporal queue of state data is embeddedly encoded to obtain the temporal queue of the target RGV car object running state embedded features and the temporal queue of the RGV preceding car running state embedded features; a temporal feature semantic transfer unit is used to perform gated feature dynamic semantic transfer on the temporal queue of the target RGV car object running state embedded features and the temporal queue of the RGV preceding car running state embedded features respectively to obtain the target RGV car object running state temporal semantic transfer representation and the RGV preceding car running state temporal semantic transfer representation; an interaction unit is used to perform running state attention interaction on the target RGV car object running state temporal semantic transfer representation and the RGV preceding car running state temporal semantic transfer representation to obtain the target RGV car object-RGV preceding car running state interaction representation, and obtain the running mode label of the target RGV car object according to the target RGV car object-RGV preceding car running state interaction representation.

[0008] According to another aspect of the present application, a circular RGV automatic control method is provided, which includes: obtaining a timing queue of running status data of a target RGV trolley object, and obtaining real-time running status data of a preceding vehicle of the target RGV trolley object on a circular track to obtain a timing queue of running status data of the preceding vehicle of the RGV; embedding and encoding the timing queue of running status data of the target RGV trolley object and the timing queue of running status data of the preceding vehicle of the RGV to obtain a timing queue of running status embedded features of the target RGV trolley object and a timing queue of running status embedded features of the preceding vehicle of the RGV; embedding and encoding the running status embedded features of the target RGV trolley object and the timing queue of running status embedded features of the preceding vehicle of the RGV The timing queue of the embedded features and the timing queue of the running state of the RGV preceding vehicle are respectively subjected to gated feature dynamic semantic transfer to obtain the running state temporal semantic transfer representation of the target RGV vehicle object and the running state temporal semantic transfer representation of the RGV preceding vehicle; the running state attention interaction is performed on the running state temporal semantic transfer representation of the target RGV vehicle object and the running state temporal semantic transfer representation of the RGV preceding vehicle to obtain the target RGV vehicle object-RGV preceding vehicle running state interaction representation, and the running mode label of the target RGV vehicle object is obtained according to the target RGV vehicle object-RGV preceding vehicle running state interaction representation.

[0009] Compared with the prior art, the circular RGV automatic control system and method provided by the present application first receives the task information sent by the RGV dispatching system and assigns it to the RGV trolley, then switches the operation mode of the RGV trolley according to the task information and the real-time status of other trolleys on the circular track, and finally instructs the RGV trolley to perform actions corresponding to the operation mode to complete the operation task. In this way, the effective dispatch and control of the RGV trolley can be realized, which is conducive to improving the efficiency of logistics transportation. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0011] In the attached picture: Figure 1 4 is a block diagram of a ring-shaped RGV automatic control system according to an embodiment of the present application.

[0012] Figure 2 4 is a block diagram of a mode switching module in a ring RGV automatic control system according to an embodiment of the present application.

[0013] Figure 3It is a block diagram of a timing feature semantic transmission unit in a ring RGV automatic control system according to an embodiment of the present application.

[0014] Figure 4 4 is a block diagram of an interaction unit in a ring RGV automatic control system according to an embodiment of the present application.

[0015] Figure 5 Flow chart of a method for automatically controlling a ring RGV according to an embodiment of the present application. DETAILED DESCRIPTION

[0016] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0017] Based on patent CN117707173A, this application proposes a ring RGV automatic control system. Figure 1 FIG. 1 is a block diagram of a ring RGV automatic control system according to an embodiment of the present application. Figure 1 As shown, in the circular RGV automatic control system 100, it includes: a task allocation module 110, which is used to receive the task information sent by the RGV scheduling system and allocate the task information to the RGV car; a mode switching module 120, which is used to switch the operation mode of the RGV car according to the task information and in combination with the real-time operation status of other RGV cars on the circular track; a car execution module 130, which is used to instruct the RGV car to perform actions corresponding to the operation mode to complete the operation task.

[0018] In an embodiment of the present application, the task assignment module 110 is used to receive task information sent by the RGV scheduling system and assign the task information to the RGV car. It should be understood that the RGV scheduling system, as the overall command center, is responsible for generating task information according to the material handling needs in the factory or warehouse, and assigning tasks to the RGV car that is most suitable for performing the task, taking into account factors such as the current position and operating status of different RGV cars (such as whether they are idle or whether they are performing other tasks). In this way, by reasonably allocating tasks, RGV car resources can be fully utilized. After receiving the task information, the RGV car can clearly know its task triggering conditions (such as where and when to start the task) and the specific requirements of the task operation (such as running in a straight line to a certain position or turning to a designated location to pick up or deliver goods), which enables the RGV car to perform tasks according to precise instructions and improves the accuracy of task execution.

[0019] The following is a detailed description of an implementable method of "receiving task information sent by the RGV scheduling system and assigning the task information to the RGV vehicle": The RGV scheduling system will be deeply integrated with the production management system, inventory management system, etc., and obtain key data such as the production progress, inventory location, and order requirements of materials in real time to determine the details of each handling task, including the type, quantity, source location, and target location of the materials to be handled. Subsequently, the scheduling system will transmit the task information to the ring RGV automatic control system through carefully selected wired or wireless communication methods. In the field of wired communications, technologies such as Ethernet and industrial fieldbus (such as Profibus, Profinet, etc.) have become common choices due to their stability and reliability; in scenarios with high flexibility requirements, wireless communication methods such as Wi-Fi, Bluetooth, ZigBee, or dedicated industrial wireless communication protocols play an important role, ensuring that task information can cross various complex environments and be accurately and timely delivered to the automatic control system.

[0020] The task assignment module in the ring RGV automatic control system receives task information through a communication interface that is compatible with the transmission mode of the dispatching system. For example, when Ethernet communication is used, the task assignment module receives data with the help of an Ethernet network card. Once the data is received, the task assignment module quickly starts the parsing program. During the parsing process, it accurately extracts key data fields from the task information, such as task number, material information, source address, target address, task priority, and estimated execution time, with advanced algorithms and logic recognition capabilities. After parsing, this information is converted into a format that can be directly processed within the system, laying a solid foundation for subsequent task assignment and execution.

[0021] In the task allocation process, the task allocation module follows a rigorous and intelligent allocation strategy. It works in collaboration with positioning systems (such as laser navigation, magnetic stripe navigation, visual navigation, etc.) to obtain the precise position information of each RGV car on the circular track in real time. Based on this, the distance between the car and the task source position and the target position is calculated, and the task is preferentially allocated to the car that is closer, thereby greatly reducing the idle time and improving the overall transportation efficiency. At the same time, the module pays close attention to the operating status of the RGV car, comprehensively considering factors such as whether the car is idle, whether it is performing other tasks, and the current power (for electric RGV cars) or energy reserve status. For example, among many candidate cars, idle and fully charged cars are given priority to perform tasks to ensure that the task can proceed smoothly and avoid delays due to insufficient energy or task conflicts. In addition, task priority is also an important consideration in allocation decisions. For those tasks with high urgency, such as the urgent need to supply raw materials to the production line, the system will break the routine and prioritize assigning them to the cart that can respond at the fastest speed, even if the cart may need to temporarily interrupt the current lower priority task (provided the system allows such operation) to ensure that production continuity is not affected.

[0022] In an embodiment of the present application, the mode switching module 120 is used to switch the operation mode of the RGV trolley according to the task information and in combination with the real-time operation status of other RGV trolleys on the circular track. It should be understood that the pickup or delivery location and the straight or turning operation requirements in the task information directly affect the initial operation mode of the RGV trolley. For example, when the task is to run in a straight line from the current position to the pickup point in area A of the warehouse, the system will set the initial operation mode of the RGV trolley to a straight-line acceleration start according to the task information, and move in the direction of area A. In other words, the task information is the basis for starting the operation mode, which determines the basic movement direction and speed state of the trolley when it starts running.

[0023] Correspondingly, in the mode switching module, the technical concept of the present application is to obtain the timing queue of the running state data of the target RGV trolley object, and obtain the real-time running state data of the preceding vehicle of the target RGV trolley object on the circular track to obtain the timing queue of the running state data of the preceding vehicle of the RGV, and use the data analysis and coding method based on artificial intelligence to embed the running state data of each target RGV trolley object and the running state data of each preceding vehicle of the RGV, respectively. Then, the gated feature semantic transfer is performed on the running state embedded features of each target RGV trolley object and the running state embedded features of each preceding vehicle of the RGV, respectively, so as to intelligently obtain the running mode label of the target RGV trolley object based on the attention interaction representation between the running state timing semantic transfer representation of the target RGV trolley object and the running state timing semantic transfer representation of the preceding vehicle of the RGV. In this way, the system can perceive the running state in real time and intelligently optimize the operation of the RGV trolley to maintain the best performance. At the same time, a more accurate and rapid response is achieved in a complex environment, the risk of misoperation is reduced, unnecessary speed reduction or parking caused by a fixed threshold is avoided, and the operating efficiency is improved.

[0024] Specifically, Figure 2 FIG. 1 is a block diagram of a mode switching module in a ring RGV automatic control system according to an embodiment of the present application. Figure 2 As shown, the mode switching module 120 includes: a vehicle operation data acquisition unit 121, which is used to acquire the timing queue of the operation status data of the target RGV vehicle object, and acquire the real-time operation status data of the preceding vehicle of the target RGV vehicle object on the circular track to obtain the timing queue of the operation status data of the preceding vehicle of the RGV; an embedding unit 122, which is used to embed the timing queue of the operation status data of the target RGV vehicle object and the timing queue of the operation status data of the preceding vehicle of the RGV to obtain the timing queue of the operation status embedded features of the target RGV vehicle object and the timing queue of the operation status embedded features of the preceding vehicle of the RGV; a timing feature semantic transmission unit 123, which is used to embed the target RGV vehicle object and the timing queue of the operation status embedded features of the preceding vehicle of the RGV The timing queue of the embedded features of the running state of the car object and the timing queue of the embedded features of the running state of the RGV preceding vehicle are respectively subjected to gated feature dynamic semantic transfer to obtain the target RGV car object running state timing semantic transfer representation and the RGV preceding vehicle running state timing semantic transfer representation; the interaction unit 124 is used to perform running state attention interaction on the target RGV car object running state timing semantic transfer representation and the RGV preceding vehicle running state timing semantic transfer representation to obtain the target RGV car object-RGV preceding vehicle running state interaction representation, and obtain the running mode label of the target RGV car object according to the target RGV car object-RGV preceding vehicle running state interaction representation.

[0025] In an embodiment of the present application, the trolley operation data acquisition unit 121 is used to acquire the timing queue of the operation status data of the target RGV trolley object, and acquire the real-time data of the operation status of the preceding vehicle of the target RGV trolley object on the circular track to obtain the timing queue of the operation status data of the preceding vehicle of the RGV. It should be understood that the timing queue of the operation status data of the target RGV trolley object contains the position information, speed information and other data of the target RGV trolley object at multiple time points. Specifically, the position information refers to the absolute position coordinates of the target RGV trolley object on the circular track, which helps to determine the current track position of the trolley and judge whether it is close to key areas such as curves, intersections or target stops. Speed ​​information refers to the running speed value of the target RGV trolley object, which can reflect the speed of the trolley. Speed ​​information is crucial for judging whether the trolley is operating normally and whether it needs to accelerate or decelerate. For example, when the target RGV car is too fast and is about to enter a curve, it may need to slow down in advance to ensure safe passage through the curve, to avoid derailment or collision with other cars due to excessive speed; on a straight track, if the speed is too slow, it may affect the overall material handling efficiency, and appropriate acceleration may be required at this time. The time series queue of the RGV front car running status data reflects the position information, speed information and other data of the front car running at multiple time points, which is the key basis for the target RGV car to adjust its own speed. For example, when the front car approaches the curve, the target RGV car can make corresponding preparations in advance according to the position information of the front car, such as slowing down or maintaining a safe distance to avoid rear-end collision or collision accidents at the curve; when the current car suddenly slows down, the target RGV car needs to sense it in time and slow down accordingly to maintain a safe distance between cars to prevent rear-end collision; when the current car accelerates and the distance between cars is large, the target RGV car can decide whether to accelerate and follow up appropriately according to the system strategy to improve transportation efficiency. In other words, the operating environment on the circular track is dynamically changing. When multiple RGV cars are running at the same time, their positions, speeds, etc. are constantly changing. The timing queue for obtaining the running status data of the target RGV vehicle object and the timing queue for obtaining the running status data of the preceding RGV vehicle can capture this dynamic change in real time, providing the system with comprehensive and accurate perception of the vehicle running status, and thus making timely and reasonable control decisions based on the real-time vehicle running status data.

[0026] The following is a detailed description of an implementable method of "obtaining a time sequence queue of the running status data of the target RGV trolley object, and obtaining the real-time running status data of the preceding vehicle of the target RGV trolley object on the circular track to obtain the time sequence queue of the running status data of the preceding vehicle of the RGV": First, for the acquisition of the running status data of the target RGV trolley to construct a time sequence queue, it is necessary to equip each RGV trolley with various types of advanced sensors. Among them, high-precision GPS sensors or indoor positioning systems (such as positioning systems based on cutting-edge technologies such as laser and ultra-wideband) play a vital role, which can determine the position information of the trolley on the circular track with extremely high accuracy. The acquisition frequency of these sensors is carefully set, usually up to 1 time per second or even higher, with the purpose of closely tracking the movement trajectory of the trolley and not missing any subtle position changes. At the same time, speed sensors are also indispensable, such as Doppler radar sensors, which are responsible for measuring the running speed of the trolley in real time. The acquisition frequency of the speed sensor is synchronized with or higher than the acquisition of the position information, because the change in speed is critical for the system to accurately grasp the running status of the trolley. Once these position and speed data are collected, they will be recorded in detail in strict chronological order, gradually forming a time series queue of the target RGV vehicle object's operating status data. At the moment the system starts, the data recording program will start synchronously, and each time a new set of data containing the current time point position and speed information is obtained, it will be accurately added to the end of the time series queue. As time goes on, data from many time points continues to accumulate, and eventually builds a sequence that can clearly reflect the trend of the vehicle's operating status changes over a period of time. This sequence provides a solid data foundation for subsequent analysis and decision-making.

[0027] In terms of obtaining the real-time running status data of the preceding vehicle of the target RGV vehicle object on the circular track to build the corresponding time sequence queue, the circular RGV automatic control system cleverly uses wireless communication technology (such as widely used Wi-Fi, stable and reliable Bluetooth, or specially designed industrial wireless communication protocols) to establish a real-time and efficient data communication link between each RGV vehicle. While focusing on obtaining its own running status data, each vehicle also has the ability to keenly receive status information from the preceding vehicle. In order to ensure accurate identification and interaction of data, the system gives each vehicle a unique identifier. For example, when an RGV vehicle starts, it quickly broadcasts its identifier and initial status information to the surrounding environment. After receiving this information, other vehicles can accurately determine the vehicles adjacent to and in front of them with the help of pre-set communication protocols and intelligent algorithms, and immediately establish a continuous and stable data receiving channel. Key running status data such as the position and speed of the preceding vehicle will be transmitted to the target RGV vehicle object in real time at a specific frequency, which is usually matched with the target vehicle's own data collection frequency, or determined according to the overall optimization requirements of the system. After receiving these data, the target car will also record and organize them in strict chronological order, thus forming a time sequence queue of the running status data of the RGV's front car. In the process of data integration, the system will use advanced timestamp synchronization processing technology to ensure that the front car data obtained by the target car is highly consistent with its own data in the time dimension, which is crucial for accurately analyzing the relative status and dynamic relationship between the two cars.

[0028] From a system level, the data transmission and storage links also have a strict and efficient guarantee mechanism. The target RGV car's own operating status data collected and the received operating status data of the preceding car will be transmitted back to the central server of the ring RGV automatic control system in real time at the fastest speed through a powerful wireless communication module. In order to ensure the stability and reliability of the data transmission process, the communication system adopts a series of advanced technical means. For example, during the data transmission process, the sender will use complex algorithms to check and calculate the data, and send the checksum together with the data to the receiver. After receiving the data, the receiver will use the same algorithm to calculate the checksum again. If the two do not match, the system will automatically trigger a mechanism to request the sender to retransmit the data to ensure the accuracy of the data. At the same time, in order to cope with possible communication link failures, the system also sets up redundant communication links. When the main link fails unexpectedly, the system can automatically and seamlessly switch to the backup link, thereby ensuring that data transmission is not interrupted and ensuring the continuous and stable operation of the entire system.

[0029] After receiving the data, the central server will properly store it in a specially designed database so that key components such as the subsequent mode switching module can obtain it in time and conduct in-depth analysis and processing. The data storage structure has been carefully optimized and designed specifically for the characteristics of time series queue data to achieve efficient storage and fast retrieval. For example, an advanced database management system based on time series is adopted. The system can quickly insert new data while supporting fast queries for data within a specific time period, and has excellent data compression capabilities, which can not only save valuable storage space but also greatly improve data processing efficiency. In addition, the system also attaches great importance to data backup and archiving management, regularly backs up historical data to external storage devices (such as tape libraries, cloud storage and other reliable storage media), and archives the data according to strict rules. This not only helps to quickly recover data when the system fails or data is lost, but also provides rich data resources for important tasks such as subsequent historical data analysis and system performance evaluation, laying a solid foundation for the continuous optimization and upgrading of the system.

[0030] In the embodiment of the present application, the embedding unit 122 is used to embed the timing queue of the running status data of the target RGV car object and the timing queue of the running status data of the preceding RGV car to obtain the timing queue of the running status embedded features of the target RGV car object and the timing queue of the running status embedded features of the preceding RGV car. Specifically, in the embodiment of the present application, the embedding unit includes: a target RGV car object running status data embedding coding subunit, which is used to use the running status embedding matrix to embed the running status data of each target RGV car object in the timing queue of the running status data of the target RGV car object to obtain the timing queue of the running status embedded coding vector of the target RGV car object as the timing queue of the running status embedded features of the target RGV car object; an RGV preceding car running status data embedding coding subunit, which is used to use the running status embedding matrix to embed the running status data of each RGV preceding car in the timing queue of the running status data of the preceding RGV car to obtain the timing queue of the running status embedded coding vector of the preceding RGV car as the timing queue of the running status embedded features of the preceding RGV car.

[0031] It should be understood that the running state data of each target RGV car object often contains information of multiple dimensions, such as speed, position and other parameters, and the amount of data will be relatively large, and the running state data at each time point contains key feature information. Therefore, in the technical solution of the present application, the running state embedding matrix is ​​used to embed the running state data of each target RGV car object in the time series queue of the running state data of the target RGV car object to map the high-dimensional data to a relatively low-dimensional space, while retaining key information as much as possible, reducing the amount of data and computational complexity of subsequent processing, and obtaining the time series queue of the target RGV car object running state embedded coding vector. In this way, the multi-dimensional running state at each time point can be converted into a vector of fixed length, which can better capture the essential characteristics of the running state of the target RGV car, so as to better understand the behavior pattern of the target RGV car over time. In particular, in the specific implementation, the running state embedding matrix is ​​usually obtained through training, which contains the mapping rules from high-dimensional space to low-dimensional space. This matrix is ​​customized according to specific task requirements to ensure that the mapped vector can effectively represent the key features of the original data.

[0032] Similarly, considering that the running state of the preceding vehicle contains a lot of information, but not all information is equally important for judging the running mode of the target RGV car, based on this, in order to extract the most critical features for mode judgment according to the characteristics of the data itself and the focus of the system, and weaken those relatively redundant or interfering information, in the technical solution of the present application, the running state embedding matrix is ​​used to embed and encode each RGV preceding vehicle running state data in the time series queue of the RGV preceding vehicle running state data to convert the multi-dimensional running state of the preceding vehicle at different time points into a vector of fixed length, and capture and mine the potential key information of the running state, and obtain the time series queue of the RGV preceding vehicle running state embedded coding vector. These vectors can more effectively represent the behavior pattern of the preceding vehicle and its changing trend, which helps to understand the impact of the preceding vehicle on the target RGV car.

[0033] In an embodiment of the present application, the temporal feature semantic transfer unit 123 is used to perform gated feature dynamic semantic transfer on the temporal queue of the target RGV trolley object running state embedded feature and the temporal queue of the RGV front vehicle running state embedded feature to obtain the target RGV trolley object running state temporal semantic transfer representation and the RGV front vehicle running state temporal semantic transfer representation. It should be understood that the temporal queue of the target RGV trolley object running state embedded coding vector and the temporal queue of the RGV front vehicle running state embedded coding vector not only cover the overall time trend, but also accommodate the running state temporal feature information presented in a short period of time, and each time point has a different degree of importance or significance in the entire time series. Therefore, in the technical solution of the present application, the gated feature dynamic semantic transfer is performed on the temporal queue of the target RGV trolley object running state embedded coding vector to obtain the target RGV trolley object running state temporal semantic transfer representation vector. In this way, the dependency relationship within a long time span can be effectively captured, and short-term changes can be quickly responded to, and the semantic association of these vectors in the time series can be deeply excavated. For example, changes in the car's speed at different points in time may indicate a transition in its operating state, such as from acceleration to deceleration or from straight-line driving to turning, helping the system better predict future behavior and make corresponding adjustments.

[0034] Specifically, Figure 3 FIG. 1 is a block diagram of a time sequence feature semantic transmission unit in a ring RGV automatic control system according to an embodiment of the present application. Figure 3 As shown, the timing feature semantic transfer unit 123 includes: a target RGV vehicle object running state feature semantic transfer sub-unit 1231, which is used to perform the gated feature dynamic semantic transfer on the timing queue of the target RGV vehicle object running state embedded coding vector to obtain the target RGV vehicle object running state timing semantic transfer representation vector as the target RGV vehicle object running state timing semantic transfer representation; an RGV preceding vehicle running state sub-unit feature semantic transfer sub-unit 1232, which is used to perform the gated feature dynamic semantic transfer on the timing queue of the RGV preceding vehicle running state embedded coding vector to obtain the RGV preceding vehicle running state timing semantic transfer representation vector as the RGV preceding vehicle running state timing semantic transfer representation.

[0035] More specifically, in the embodiment of the present application, the target RGV trolley object running state feature semantic transmission subunit includes: a target RGV trolley object running state feature dynamic propagation secondary subunit, which is used to perform a forward LSTM model sequence propagation on the time series queue of the target RGV trolley object running state embedded coding vector to obtain the initial target RGV trolley object running state feature dynamic propagation coding vector; a target RGV trolley running state private gated feature calculation secondary subunit, which is used to use the initial target RGV trolley object running state feature dynamic propagation coding vector as an aggregate representation to calculate the target RGV trolley object running state feature dynamic propagation coding vector. The target RGV car object running state embedded coding vectors in the timing queue of the car object running state embedded coding vector are relative to the private gating features of the aggregate representation to obtain the timing queue of the gated writing target RGV car object running state private feature coding vector; the target RGV car object running state timing semantic transfer representation generates a secondary sub-unit, which is used to fuse the timing queue of the gated writing target RGV car object running state private feature coding vector and the initial target RGV car object running state feature dynamic propagation coding vector to obtain the target RGV car object running state timing semantic transfer representation vector.

[0036] Specifically, first, the sequence of the target RGV object running state embedding coding vector is forward propagated by the LSTM model to obtain the initial target RGV object running state feature dynamic propagation coding vector. The above process can be expressed as: ;in, is the time sequence queue of the target RGV object running state embedded coding vector, and They are the first, second, and third in the time sequence queue of the target RGV object running state embedded coding vector. and The target RGV vehicle object running state embedding coding vector, is the forward LSTM encoding, It is the dynamic propagation coding vector of the running state characteristics of the initial target RGV vehicle object.

[0037] It should be understood that as a variant of recurrent neural network (RNN), forward LSTM can effectively capture long-range dependencies in sequence data. By performing sequence propagation of the forward LSTM model on the time series queue of the target RGV car object running state embedding encoding vector, the model selectively updates, retains and outputs information through its internal gating mechanism (input gate, forget gate, output gate), so as to accurately learn the change pattern of the target RGV car object running state at different time points, and then grasp the detailed characteristics of the evolution of the overall running state of the car, and generate a hidden state representation that integrates the target RGV car object running state information at all previous time points, that is, the initial target RGV car object running state feature dynamic propagation encoding vector.

[0038] Specifically, in an embodiment of the present application, the target RGV car running state private gated feature calculation secondary subunit includes: a target RGV car running state private feature calculation tertiary subunit, which is used to calculate the private features of each target RGV car object running state embedded coding vector in the timing queue of the target RGV car object running state embedded coding vector relative to the initial target RGV car object running state feature dynamic propagation coding vector to obtain the timing queue of the target RGV car object running state private feature coding vector; a target RGV car running state private feature gated write tertiary subunit, which is used to perform gated write based on the Softmax function on each target RGV car object running state private feature coding vector in the timing queue of the target RGV car object running state private feature coding vector to obtain the timing queue of the gated write target RGV car object running state private feature coding vector.

[0039] More specifically, in the embodiment of the present application, the target RGV car running state private feature calculation three-level subunit is used to: calculate the position point subtraction between the target RGV car object running state embedded coding vector and the initial target RGV car object running state feature dynamic propagation coding vector, and then perform probabilistic point convolution and Sigmoid function on the obtained target RGV car object running state difference vector to obtain the target RGV car object running state weight vector. The process can be expressed as: ;in, is the dynamic propagation coding vector of the running state characteristics of the initial target RGV vehicle object, It is the first in the time sequence queue of the target RGV object running state embedded coding vector. The target RGV vehicle object running state embedding coding vector, It is subtracted by position point. Indicates taking the absolute value, is the point convolutional coding, is the weight matrix, is the activation function, yes The corresponding target RGV vehicle object running state weight vector.

[0040] Based on the target RGV trolley object running state weight vector, the target RGV trolley object running state embedded coding vector is weighted to obtain the target RGV trolley object running state private feature coding vector. This process can be expressed as: ;in, It is the first in the time sequence queue of the target RGV object running state embedded coding vector. The target RGV vehicle object running state embedding coding vector, is the point convolutional coding, is the activation function, yes The corresponding target RGV vehicle object running state weight vector, It is the first in the time sequence queue of the private feature encoding vector of the target RGV object's running state. The private feature encoding vector of the running status of the target RGV vehicle object.

[0041] It should be understood that the initial target RGV car object running state feature dynamic propagation coding vector is obtained after the forward LSTM model performs comprehensive processing on the entire time series queue, which focuses on reflecting the overall and common running state change trend and comprehensive information. However, in this integration process, the target RGV car object running state embedding coding vector corresponding to each individual time point actually contains some unique detail information, which may not be focused on in the sequence propagation. In order to mine the local subtle information that is easily overlooked, so as to have a more nuanced understanding of the target RGV car running state, it is necessary to calculate the private features of each target RGV car object running state embedding coding vector in the time series queue of the target RGV car object running state embedding coding vector relative to the initial target RGV car object running state feature dynamic propagation coding vector. Through this processing, a series of target RGV car object running state private feature coding vectors that emphasize the individual differences of the car running state at each time point can be generated.

[0042] Then, each target RGV trolley object running state private feature coding vector in the time series queue of the target RGV trolley object running state private feature coding vector is gated and written based on the Softmax function to obtain the time series queue of the gated target RGV trolley object running state private feature coding vector. The above process can be expressed as: ;in, It is the first in the time sequence queue of the private feature encoding vector of the target RGV object's running state. The private feature encoding vector of the running state of the target RGV vehicle object, is matrix multiplication, yes The corresponding target RGV car object running state private feature weight matrix, yes The corresponding target RGV car object running state private feature bias vector, yes The node position in the time series queue of the target RGV vehicle object running state private feature encoding vector, yes function, It is the first in the time sequence queue of the private feature activation vector of the target RGV object running state. The target RGV car object running state private feature activation vector, For Perform mask operations. yes The eigenvalues ​​at each position in , is the preset threshold, It is the first in the time sequence queue of the private feature mask vector of the target RGV object's running state. The private feature mask vector of the running state of the target RGV car object, and are the first, second and third in the time series queue of the private feature encoding vector of the target RGV object's running state. The private feature encoding vector of the running state of the target RGV vehicle object, and are the first, second and third in the time series queue of the private feature mask vector of the target RGV object's running state. The private feature mask vector of the running state of the target RGV car object, It is a time sequence queue for gated writing of the private feature encoding vector of the target RGV vehicle object's running status.

[0043] It should be understood that when the gated write operation based on the Softmax function is performed on the timing queue of the private feature encoding vector of the target RGV car object's running state, it draws on the principle of synaptic plasticity in biology. This method of simulating the biological information processing mechanism can add rationality to the entire information screening process. Specifically, when analyzing the target RGV car running state information, the gated write mechanism can reasonably determine which information is critical for subsequent analysis and which is relatively minor based on the importance of the features reflected by the private feature encoding vectors of the running state of each target RGV car object, so that the information contained in the timing queue obtained after the gated write is strictly screened and the key content that is most helpful for grasping the core state of the car's operation. That is, the gated write operation based on the Softmax function is used to determine whether the private features of the running state of each target RGV car object need to be written into the initial propagation reasoning information of the feature sequence.

[0044] Finally, the time sequence queue of the gated writing target RGV trolley object running state private feature encoding vector and the initial target RGV trolley object running state feature dynamic propagation encoding vector are merged to obtain the target RGV trolley object running state time sequence semantic transfer representation vector. The above process can be expressed as: ;in, is the dynamic propagation coding vector of the running state characteristics of the initial target RGV vehicle object, It is a time sequence queue for gated writing of the private feature encoding vector of the target RGV vehicle object's running status. For cascade processing, It is the temporal semantic transfer representation vector of the target RGV vehicle object's running status.

[0045] That is, by fusing the time series queue of the gated writing target RGV car object running state private feature coding vector and the initial target RGV car object running state feature dynamic propagation coding vector, it can help to better understand how the running states of the target RGV car object at various time points in the entire running cycle influence and restrict each other, that is, it can clearly show the dynamic interaction mode of the car running state evolving over time. At the same time, the generated target RGV car object running state time series semantic transfer representation vector can also accurately capture the unique key information of the car running state at each time point.

[0046] In the embodiment of the present application, the interaction unit 124 is used to perform operation state attention interaction on the target RGV vehicle object operation state temporal semantic transfer representation and the RGV front vehicle operation state temporal semantic transfer representation to obtain the target RGV vehicle object-RGV front vehicle operation state interaction representation, and obtain the operation mode label of the target RGV vehicle object according to the target RGV vehicle object-RGV front vehicle operation state interaction representation. Specifically, Figure 4 FIG. 1 is a block diagram of an interactive unit in a ring RGV automatic control system according to an embodiment of the present application. Figure 4 As shown, the interaction unit 124 includes: a target RGV car object-RGV preceding vehicle running state interaction representation generating subunit 1241, which is used to perform running state cross-attention interaction on the target RGV car object running state temporal semantic transfer representation vector and the RGV preceding vehicle running state temporal semantic transfer representation vector to obtain a target RGV car object-RGV preceding vehicle running state interaction representation vector as the target RGV car object-RGV preceding vehicle running state interaction representation; a target RGV car object running mode generating subunit 1242, which is used to obtain a mode switching result according to the target RGV car object-RGV preceding vehicle running state interaction representation vector, and the mode switching result is used to represent the running mode label of the target RGV car object.

[0047] In an embodiment of the present application, the target RGV trolley object-RGV preceding vehicle running state interaction representation generating subunit 1241 is used to perform running state cross-attention interaction on the target RGV trolley object running state temporal semantic transfer representation vector and the RGV preceding vehicle running state temporal semantic transfer representation vector to obtain the target RGV trolley object-RGV preceding vehicle running state interaction representation vector as the target RGV trolley object-RGV preceding vehicle running state interaction representation. It should be understood that the target RGV trolley object running state temporal semantic transfer representation and the RGV preceding vehicle running state temporal semantic transfer representation influence each other. In the circular track running scenario, the running mode of the target trolley (such as acceleration, deceleration, turning, etc.) needs to fully consider the state of the preceding vehicle. Therefore, in order to more accurately capture the mutual influence and synergy between the two vehicles, so as to provide more detailed and comprehensive information support for intelligent decision-making, in the technical solution of the present application, the target RGV car object running state temporal semantic transfer representation vector and the RGV front car running state temporal semantic transfer representation vector are subjected to running state cross-attention interaction to obtain the target RGV car object-RGV front car running state interaction representation vector. It should be explained that cross attention is an attention mechanism, which is mainly used to deal with the correlation between two or more different information sequences. Simply put, the attention mechanism is like a spotlight, which can focus on certain important parts in complex information. Cross attention is to switch the spotlight back and forth between two different information sets, find the interrelated parts between them, and emphasize the importance of these related parts. Specifically, first, the target RGV car object running state temporal semantic transfer representation vector and the RGV front car running state temporal semantic transfer representation vector are linearly changed respectively to obtain the query vector, key vector and value vector corresponding to the running state of the target RGV car object and the query vector, key vector and value vector corresponding to the running state of the RGV front car. Then, the query vector corresponding to the target RGV car object running state, the key vector and value vector corresponding to the RGV front car running state, and the key vector, value vector corresponding to the target RGV car object running state and the query vector corresponding to the RGV front car running state are input into the Transformer to capture the dynamic interaction mode between the target RGV car and its front car, and obtain the target RGV car-front car running state interaction vector and the front car-target RGV car running state interaction vector. Then, these two interaction vectors are fused to obtain the target RGV car object-RGV front car running state interaction representation vector containing the association information of the two cars.

[0048] In an embodiment of the present application, the target RGV trolley object operation mode generation subunit 1242 is used to obtain a mode switching result according to the target RGV trolley object-RGV preceding vehicle operation state interaction characterization vector, and the mode switching result is used to represent the operation mode label of the target RGV trolley object. Specifically, in an embodiment of the present application, the target RGV trolley object operation mode generation subunit is used to: pass the target RGV trolley object-RGV preceding vehicle operation state interaction characterization vector through a classifier-based control mode switcher to obtain the mode switching result, and the mode switching result is used to represent the operation mode label of the target RGV trolley object. It should be understood that the target RGV trolley object-RGV preceding vehicle operation state interaction characterization vector integrates the target RGV trolley's own operation state and the complex interaction relationship information with the preceding vehicle. In order to convert the rich information in this vector into specific operation instructions, it is necessary to input the target RGV trolley object-RGV preceding vehicle operation state interaction characterization vector into the classifier-based control mode switcher. The classifier specifically refers to a machine learning model that can map the interaction representation vector to a specific operation mode label according to predefined rules and learned patterns, thereby achieving precise control of the operation mode of the RGV car. The operation mode labels in this application mainly include linear acceleration, linear uniform speed, linear deceleration, turning (specific turning modes for curves with different curvatures), parking, etc. These labels clarify the basic movement mode of the target RGV car on the circular track and are the most direct instructions for the system to control the operation of the car. The mode switching result obtained by classification directly determines the next action of the target RGV car, which enables the RGV car to make accurate action responses according to the actual operating environment and task requirements.

[0049] In particular, when the target RGV car object running state temporal semantic transfer representation vector and the RGV preceding car running state temporal semantic transfer representation vector respectively represent the embedded coding temporal dynamic propagation characteristics of the running state data of the target RGV car and its preceding car, when performing RGV running state interaction encoding based on the cross-attention module, the source sample differentiated cross-attention misalignment will also cause the running state interaction richness of the target RGV car object-RGV preceding car running state interaction representation vector, resulting in the repetition of unstructured probability mapping, affecting the accuracy of the mode switching results.

[0050] In a preferred example, the target RGV car object-RGV preceding vehicle running state interaction characterization vector is passed through a classifier-based control mode switcher to obtain a mode switching result, including: determining the number of non-zero eigenvalues ​​in the target RGV car object-RGV preceding vehicle running state interaction characterization vector, and subtracting the number of non-zero eigenvalues ​​from the length of the target RGV car object-RGV preceding vehicle running state interaction characterization vector to obtain the target RGV car object-RGV preceding vehicle running state interaction zero-dimensional value .

[0051] Calculate the target RGV car object-RGV front car running state interaction zero-dimensional value minus one of the target RGV car object-RGV front car running state interaction zero differentiable value The first target RGV car object-RGV front vehicle running state interaction field fitting value and the second target RGV car object-RGV front vehicle running state interaction field fitting value obtained by multiplying and dividing by the target RGV car object-RGV front vehicle running state interaction zero-dimensional value respectively can be expressed as: ;in, is the target RGV vehicle object-RGV front vehicle operation state interaction zero-dimensional value, is the first target RGV vehicle object-RGV front vehicle running state interaction field fitting value, It is the fitting value of the interaction field between the second target RGV vehicle object and the RGV front vehicle running state.

[0052] The square root of the sum of squares of all eigenvalues ​​of the target RGV vehicle object-RGV preceding vehicle running state interaction characterization vector is calculated to obtain the target RGV vehicle object-RGV preceding vehicle running state interaction mode characterization value. The process can be expressed as: ;in, represents the first The eigenvalues ​​at the positions, Represents the target RGV vehicle object-RGV front vehicle operation status interaction mode representation value.

[0053] The target RGV vehicle object-RGV front vehicle operation state interaction zero differentiable value The power function of each eigenvalue of the target RGV car object-RGV preceding vehicle running state interaction characterization vector is calculated as an exponent, and multiplied by the first target RGV car object-RGV preceding vehicle running state interaction field fitting value to obtain the target RGV car object-RGV preceding vehicle running state interaction local representation value. The process can be expressed as: ;in, is the target RGV vehicle object-RGV front vehicle operation state interaction zero-dimensional value, is the first target RGV vehicle object-RGV front vehicle running state interaction field fitting value, express The corresponding target RGV vehicle object-RGV front vehicle running status interaction local representation value.

[0054] Calculate each eigenvalue of the target RGV vehicle object-RGV front vehicle running state interaction representation vector The product of the second target RGV car object-RGV front vehicle running state interaction field fitting value and the target RGV car object-RGV front vehicle running state interaction mode characterization value is obtained to obtain the target RGV car object-RGV front vehicle running state interaction center pointing value. The process can be expressed as: ;in, represents the target RGV vehicle object-RGV front vehicle operation state interaction mode representation value, is the fitting value of the interaction field between the second target RGV vehicle object and the RGV front vehicle running state, represents the first The eigenvalues ​​at the positions, express The corresponding target RGV vehicle object - RGV front vehicle running status interaction center pointing value.

[0055] The weighted difference between the target RGV car object-RGV front vehicle running state interaction local representation value and the target RGV car object-RGV front vehicle running state interaction center pointing value is used to form an optimized target RGV car object-RGV front vehicle running state interaction representation vector. This process can be expressed as: ;in, Indicates the local representation value of the target RGV vehicle object-RGV front vehicle operation status interaction, Indicates the target RGV vehicle object-RGV front vehicle running status interaction center pointing value, and represents the weighted hyperparameter, represents the first The feature value at each position.

[0056] The optimized target RGV vehicle object-RGV front vehicle operation state interaction representation vector is passed through the classifier-based control mode switcher to obtain the mode switching result.

[0057] That is, based on the vector field fitting law of high-dimensional manifold, the isolated zeros in the target RGV car object-RGV preceding vehicle running state interaction representation vector are used as the differentiable dimension of the high-dimensional manifold, so as to fix the normal direction pointing to the distribution center near the local coordinates represented by the eigenvalues ​​of the target RGV car object-RGV preceding vehicle running state interaction representation vector, thereby avoiding the invalid repetition in the probability mapping process caused by the unstructured characteristics of the semantically rich distribution of the target RGV car object-RGV preceding vehicle running state interaction representation vector through the effective alignment of the characteristic distribution pattern of the target RGV car object-RGV preceding vehicle running state interaction representation vector to the probability density differentiable field, thereby improving the accuracy of the mode switching result obtained by the classifier-based control mode switcher.

[0058] In summary, the mode switching module 120 is clearly explained, which uses artificial intelligence-based data analysis and coding methods to embed and encode the running status data of each target RGV car object and the running status data of each RGV front car, and then, respectively, perform gated feature semantic transfer on the running status embedded features of each target RGV car object and the running status embedded features of each RGV front car, so as to intelligently obtain the running mode label of the target RGV car object based on the attention interaction representation between the running status temporal semantic transfer representation of the target RGV car object and the running status temporal semantic transfer representation of the RGV front car. In this way, the system can perceive the running status in real time and intelligently optimize the operation of the RGV car to maintain optimal performance. At the same time, more accurate and rapid responses are achieved in complex environments, the risk of misoperation is reduced, and unnecessary speed reduction or parking caused by fixed thresholds is avoided, which helps to improve operating efficiency.

[0059] In the embodiment of the present application, the trolley execution module 130 is used to instruct the RGV trolley to perform actions corresponding to the operation mode to complete the operation task. It should be understood that according to the requirements of the operation mode, the speed can be adjusted by controlling the drive motor of the RGV trolley. For example, in the linear acceleration mode, a signal to increase the voltage or current is sent to the drive motor to make it output a greater torque, thereby driving the trolley to accelerate forward; in the deceleration mode, the power output of the drive motor is reduced or a reverse braking force is applied to gradually decelerate the trolley. For the turning operation mode, the control system sends instructions to the steering mechanism of the RGV trolley to control the steering angle of the wheel. For example, before entering the curve, the appropriate steering angle is calculated according to the curvature of the curve and the speed of the trolley, and the direction of the wheel is accurately adjusted through the steering motor or the hydraulic steering system so that the trolley can travel along the predetermined curve trajectory. Through reasonable operation mode switching and action execution, the mutual waiting time and unnecessary driving paths between RGV trolleys are reduced, and the material transportation efficiency of the entire circular track system is improved. That is, when multiple vehicles are running at the same time, each vehicle accelerates, decelerates, turns, and performs other actions in an orderly manner according to real-time operating mode instructions, avoiding traffic jams and waste of resources, and enabling the entire system to complete material handling tasks more quickly.

[0060] In summary, the circular RGV automatic control system 100 based on the embodiment of the present application is explained, which first receives the task information sent by the RGV scheduling system and assigns it to the RGV trolley, then switches the operation mode of the RGV trolley according to the task information and the real-time status of other trolleys on the circular track, and finally instructs the RGV trolley to perform actions corresponding to the operation mode to complete the operation task. In this way, effective scheduling and control of the RGV trolley can be achieved, which is conducive to improving the efficiency of logistics transportation.

[0061] Figure 5 FIG. 1 is a flow chart of a method for automatically controlling a ring RGV according to an embodiment of the present application. Figure 5As shown, in the circular RGV automatic control method, it includes: S110, obtaining the timing queue of the running status data of the target RGV trolley object, and obtaining the real-time running status data of the preceding vehicle of the target RGV trolley object on the circular track to obtain the timing queue of the running status data of the preceding vehicle of the RGV; S120, embedding the timing queue of the running status data of the target RGV trolley object and the timing queue of the running status data of the preceding vehicle of the RGV to obtain the timing queue of the running status embedded feature of the target RGV trolley object and the timing queue of the running status embedded feature of the preceding vehicle of the RGV; S130, embedding the running status of the target RGV trolley object into the timing queue of the running status embedded feature of the preceding vehicle of the RGV The timing queue of the feature and the timing queue of the RGV preceding vehicle running state embedded feature are respectively subjected to gated feature dynamic semantic transfer to obtain the target RGV vehicle object running state timing semantic transfer representation and the RGV preceding vehicle running state timing semantic transfer representation; S140, the running state attention interaction is performed on the target RGV vehicle object running state timing semantic transfer representation and the RGV preceding vehicle running state timing semantic transfer representation to obtain the target RGV vehicle object-RGV preceding vehicle running state interaction representation, and the running mode label of the target RGV vehicle object is obtained according to the target RGV vehicle object-RGV preceding vehicle running state interaction representation.

[0062] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned ring RGV automatic control method have been described in detail above. Figures 1 to 4 The description of the ring RGV automatic control system has been described in detail, and therefore, its repeated description will be omitted.

[0063] In summary, the automatic control method of the circular RGV based on the embodiment of the present application is explained, which first receives the task information sent by the RGV scheduling system and assigns it to the RGV trolley, then switches the operation mode of the RGV trolley according to the task information and the real-time status of other trolleys on the circular track, and finally instructs the RGV trolley to perform actions corresponding to the operation mode to complete the operation task. In this way, the effective scheduling and control of the RGV trolley can be realized, which is conducive to improving the efficiency of logistics transportation.

Claims

1. A ring RGV automatic control system, comprising: A task allocation module is used to receive task information sent by the RGV scheduling system and allocate the task information to the RGV vehicle; A mode switching module is used to switch the operation mode of the RGV trolley according to the task information and in combination with the real-time operation status of other RGV trolleys on the circular track; a trolley execution module is used to instruct the RGV trolley to execute actions corresponding to the operation mode to complete the operation task; it is characterized in that the mode switching module includes: a trolley operation data acquisition unit, which is used to acquire the timing queue of the operation status data of the target RGV trolley object, and acquire the real-time operation status data of the preceding vehicle of the target RGV trolley object on the circular track to obtain the timing queue of the operation status data of the preceding vehicle of the RGV; an embedding unit is used to embed the timing queue of the operation status data of the target RGV trolley object and the timing queue of the operation status data of the preceding vehicle of the RGV to obtain the target RGV trolley object. A timing queue and a timing queue with embedded features of the running status of the RGV preceding vehicle; a timing feature semantic transfer unit, used to perform gated feature dynamic semantic transfer on the timing queue with embedded features of the running status of the target RGV vehicle object and the timing queue with embedded features of the running status of the RGV preceding vehicle, respectively, to obtain the timing semantic transfer representation of the running status of the target RGV vehicle object and the timing semantic transfer representation of the running status of the RGV preceding vehicle; an interaction unit, used to perform running status attention interaction on the timing semantic transfer representation of the running status of the target RGV vehicle object and the timing semantic transfer representation of the running status of the RGV preceding vehicle to obtain the target RGV vehicle object-RGV preceding vehicle running status interaction representation, and obtain the running mode label of the target RGV vehicle object according to the target RGV vehicle object-RGV preceding vehicle running status interaction representation.

2. The ring RGV automatic control system according to claim 1 is characterized in that: The embedding unit includes: a target RGV vehicle object running status data embedding coding subunit, which is used to use the running status embedding matrix to embed the running status data of each target RGV vehicle object in the timing queue of the running status data of the target RGV vehicle object to obtain a timing queue of target RGV vehicle object running status embedded coding vectors as the timing queue of the target RGV vehicle object running status embedded features; an RGV preceding vehicle running status data embedding coding subunit, which is used to use the running status embedding matrix to embed the running status data of each RGV preceding vehicle in the timing queue of the RGV preceding vehicle running status data to obtain a timing queue of RGV preceding vehicle running status embedded coding vectors as the timing queue of the RGV preceding vehicle running status embedded features.

3. The ring RGV automatic control system according to claim 2 is characterized in that: The timing feature semantic transfer unit includes: a target RGV vehicle object running state feature semantic transfer sub-unit, which is used to perform the gated feature dynamic semantic transfer on the timing queue of the target RGV vehicle object running state embedded coding vector to obtain the target RGV vehicle object running state timing semantic transfer representation vector as the target RGV vehicle object running state timing semantic transfer representation; an RGV preceding vehicle running state sub-unit feature semantic transfer sub-unit, which is used to perform the gated feature dynamic semantic transfer on the timing queue of the RGV preceding vehicle running state embedded coding vector to obtain the RGV preceding vehicle running state timing semantic transfer representation vector as the RGV preceding vehicle running state timing semantic transfer representation.

4. The ring RGV automatic control system according to claim 3 is characterized in that: The target RGV trolley object running state feature semantic transmission subunit includes: a target RGV trolley object running state feature dynamic propagation secondary subunit, which is used to perform a forward LSTM model sequence propagation on the time series queue of the target RGV trolley object running state embedded coding vector to obtain the initial target RGV trolley object running state feature dynamic propagation coding vector; a target RGV trolley running state private gated feature calculation secondary subunit, which is used to use the initial target RGV trolley object running state feature dynamic propagation coding vector as an aggregate representation to calculate the target RGV trolley object running state The embedded coding vectors of the running states of each target RGV car object in the timing queue of the embedded coding vector are relative to the private gating features of the aggregate representation to obtain the timing queue of the gated writing of the target RGV car object running state private feature coding vector; the target RGV car object running state timing semantic transfer representation generates a secondary sub-unit, which is used to fuse the timing queue of the gated writing of the target RGV car object running state private feature coding vector and the initial target RGV car object running state feature dynamic propagation coding vector to obtain the target RGV car object running state timing semantic transfer representation vector.

5. The ring RGV automatic control system according to claim 4 is characterized in that: The target RGV car running state private gated feature calculation secondary sub-unit includes: a target RGV car running state private feature calculation tertiary sub-unit, which is used to calculate the private features of each target RGV car object running state embedded coding vector in the timing queue of the target RGV car object running state embedded coding vector relative to the initial target RGV car object running state feature dynamic propagation coding vector to obtain the timing queue of the target RGV car object running state private feature coding vector; a target RGV car running state private feature gated writing tertiary sub-unit, which is used to perform gated writing based on the Softmax function on each target RGV car object running state private feature coding vector in the timing queue of the target RGV car object running state private feature coding vector to obtain the timing queue of the gated writing target RGV car object running state private feature coding vector.

6. The ring RGV automatic control system according to claim 5, characterized in that: The target RGV car running state private feature calculation three-level sub-unit is used to: calculate the position point subtraction between the target RGV car object running state embedded coding vector and the initial target RGV car object running state feature dynamic propagation coding vector, and then perform probabilistic point convolution and Sigmoid function on the obtained target RGV car object running state difference vector to obtain the target RGV car object running state weight vector; based on the target RGV car object running state weight vector, perform weighted processing on the target RGV car object running state embedded coding vector to obtain the target RGV car object running state private feature coding vector.

7. The ring RGV automatic control system according to claim 6, characterized in that: The interaction unit includes: a target RGV car object-RGV preceding car running state interaction representation generating subunit, which is used to perform running state cross-attention interaction on the target RGV car object running state temporal semantic transfer representation vector and the RGV preceding car running state temporal semantic transfer representation vector to obtain a target RGV car object-RGV preceding car running state interaction representation vector as the target RGV car object-RGV preceding car running state interaction representation; a target RGV car object running mode generating subunit, which is used to obtain a mode switching result according to the target RGV car object-RGV preceding car running state interaction representation vector, and the mode switching result is used to represent the running mode label of the target RGV car object.

8. The ring RGV automatic control system according to claim 7, characterized in that: The target RGV vehicle object operation mode generation subunit is used to: pass the target RGV vehicle object-RGV front vehicle operation state interaction representation vector through a classifier-based control mode switcher to obtain the mode switching result, and the mode switching result is used to represent the operation mode label of the target RGV vehicle object.

9. A method for automatic control of a ring RGV, characterized in that: include: The timing queue of the running status data of the target RGV trolley object is obtained, and the real-time running status data of the preceding vehicle of the target RGV trolley object on the circular track is obtained to obtain the timing queue of the running status data of the preceding vehicle of the RGV; the timing queue of the running status data of the target RGV trolley object and the timing queue of the running status data of the preceding vehicle of the RGV are embedded and encoded to obtain the timing queue of the running status embedded features of the target RGV trolley object and the timing queue of the running status embedded features of the preceding vehicle of the RGV; the timing queue of the running status embedded features of the target RGV trolley object and the timing queue of the running status embedded features of the preceding vehicle of the RGV are embedded and encoded to obtain the timing queue of the running status embedded features of the target RGV trolley object and the timing queue of the running status embedded features of the preceding vehicle of the RGV The temporal queue of the running status embedded feature performs gated feature dynamic semantic transfer to obtain the target RGV car object running status temporal semantic transfer representation and the RGV front car running status temporal semantic transfer representation; the running status attention interaction is performed on the target RGV car object running status temporal semantic transfer representation and the RGV front car running status temporal semantic transfer representation to obtain the target RGV car object-RGV front car running status interaction representation, and according to the target RGV car object-RGV front car running status interaction representation, the running mode label of the target RGV car object is obtained.

10. The method for automatically controlling the ring RGV according to claim 9, characterized in that: Embedding and encoding the timing queue of the running status data of the target RGV vehicle object and the timing queue of the running status data of the preceding RGV vehicle to obtain the timing queue of the running status embedded features of the target RGV vehicle object and the timing queue of the running status embedded features of the preceding RGV vehicle, including: using the running status embedding matrix to embed the running status data of each target RGV vehicle object in the timing queue of the running status data of the target RGV vehicle object to obtain the timing queue of the running status embedded coding vector of the target RGV vehicle object as the timing queue of the running status embedded features of the target RGV vehicle object; using the running status embedding matrix to embed the running status data of each RGV preceding vehicle in the timing queue of the running status data of the preceding RGV vehicle to obtain the timing queue of the running status embedded coding vector of the preceding RGV vehicle as the timing queue of the running status embedded features of the preceding RGV vehicle.

Images