Method for dynamically adapting to piece feeding of large and small grids

Abstract

The embodiment of the invention discloses a method for dynamically adapting to piece feeding of large and small grids, and relates to the technical field of automatic control. The method comprises the following steps: acquiring weight data corresponding to a target package, and acquiring a plurality of package images for the target package from a plurality of different visual angles; obtaining volume features and attitude features of a target package based on the plurality of package images; matching a delivery cell corresponding to the target package, and obtaining corresponding delivery cell information; and determining a delivery control strategy corresponding to the target package according to the delivery cell information and weight data, volume characteristics and attitude characteristics corresponding to the target package, the discharging control strategy comprises a first control parameter for controlling a sorting loop line to move in the annular direction, a second control parameter for controlling a sorting platform where the target package is located to drive the target package to move towards the discharging grid opening, and a third control parameter for controlling a deviation rectifying system to rectify deviation of the target package.

Description

Technical Field

[0001] This application relates to the field of automatic control technology, and more specifically, to a method for dynamically adapting the placement of parts in slots of varying sizes. Background Technology

[0002] During parcel sorting, parcels on the sorting loop need to be sorted into different sized slots (e.g., 750 slots, 950 slots, 1300 chute, etc.) based on their size and / or weight. If a static unloading method is used (i.e., the sorting loop is briefly stopped during unloading), sorting efficiency will be significantly reduced (especially when handling large volumes of parcels, as static unloading cannot meet the demands of rapid sorting). If a dynamic unloading method is used (i.e., the sorting loop does not stop during unloading), improper control of the relative movement between the parcel and the slot may cause parcels to fail to fall accurately into the unloading slot, leading to sorting errors (especially when handling parcels of different sizes, the likelihood of sorting errors will further increase with dynamic unloading).

[0003] Therefore, it is necessary to study a method for dynamically adapting to different sizes of compartments to ensure that packages of different sizes and weights can be accurately and efficiently placed into the corresponding compartments without stopping the sorting loop. Summary of the Invention

[0004] To achieve the above objectives, embodiments of this specification provide a method for dynamically adapting parts to different grid sizes, the method comprising: The weight data of the target package and multiple images of the package were acquired from multiple different perspectives. The volume and orientation features of the target package are obtained based on the multiple package images; Match the lower compartment corresponding to the target package and obtain the corresponding lower compartment information, wherein the lower compartment information is used to describe the distribution location and compartment specifications of the lower compartment; Based on the unloading slot information and the weight, volume, and posture characteristics of the target package, an unloading control strategy for the target package is determined. The unloading control strategy includes a first control parameter for controlling the sorting loop to move in a circumferential direction, a second control parameter for controlling the sorting platform containing the target package to drive the target package toward the unloading slot, and a third control parameter for controlling the correction system to correct the deviation of the target package.

[0005] In some embodiments, obtaining the volume and orientation features of the target package based on the multiple package images includes: The multiple package images are preprocessed, and the preprocessing includes at least denoising and scaling. The denoising includes removing background information from each package image except for the target package and the sorting platform it is located on. The preprocessed images of the package are input into the trained volume recognition network and pose recognition network to extract the volume feature vector and pose feature vector corresponding to each image of the package through the volume recognition network and pose recognition network. The volume feature vectors and pose feature vectors corresponding to the multiple package images are integrated to obtain the comprehensive volume vector and comprehensive pose vector corresponding to the target package. The comprehensive volume vector and the comprehensive attitude vector are mapped to obtain the volume features and attitude features corresponding to the target package; wherein, the volume features include the maximum cross-sectional size and the minimum cross-sectional size corresponding to the target package, and the attitude features include the center of gravity position corresponding to the target package and the relative positional relationship between the target package and the sorting platform on which it is located.

[0006] In some embodiments, removing background information from each package image, excluding the target package and its sorting platform, includes: The edge detection algorithm identifies the outline of the target package corresponding to the target package in each package image, as well as the outline of the sorting platform that intersects with the outline of the target package. The edge detection algorithm includes the Canny edge detection algorithm or the Sobel edge detection algorithm. A mask image is generated based on the outline of the target package and the outline of the sorting platform, wherein the mask image retains only the image information corresponding to the target package and the sorting platform it is located in; The mask image is used to remove background information from the corresponding original package images in the multiple package images.

[0007] In some embodiments, matching the lower compartment corresponding to the target package includes: The target unloading area is determined based on the barcode information corresponding to the target package, wherein the target unloading area includes multiple compartments of different sizes; Based on the volume characteristics, the compartment with the closest size among the multiple compartments contained in the target unloading area is determined as the unloading compartment, wherein the minimum channel size corresponding to the unloading compartment is greater than the maximum cross-sectional size corresponding to the target package.

[0008] In some embodiments, determining the unloading control strategy corresponding to the target package based on the unloading compartment information and the weight data, volume characteristics, and posture characteristics corresponding to the target package includes: The information of the lower compartment, as well as the weight data, volume features, and posture features of the target package, are used as input data to input into the trained lower control strategy generation model, so that the lower control strategy generation model can output a lower control strategy that matches the target package. The lower control strategy generation model includes a neural network model, which includes an input layer, an output layer, and multiple hidden layers. The input layer is used to receive the input data, and the output layer is used to output the first control parameter, the second control parameter, and the third control parameter.

[0009] In some embodiments, the first control parameter includes the movement speed, acceleration, and deceleration point of the sorting loop in the unloading stage; the second control parameter includes the driving force, driving speed, and driving direction of the sorting platform in the unloading stage; and the third control parameter includes the correction force and correction angle of the correction system.

[0010] In some embodiments, the lower-component control strategy generation model is trained based on the following method: A training sample set is constructed, which includes multiple training sample data. Each training sample data includes the lower part compartment information, weight data, volume features, posture features, and a lower part control strategy label corresponding to the sample package. The lower part control strategy label includes a first sample control parameter, a second sample control parameter, a third sample control parameter, and a corresponding control result score. The first sample control parameter, the second sample control parameter, the third sample control parameter, and the corresponding control result score are obtained based on experimental testing. The initial neural network model is trained using the training sample set, and the parameters of the initial neural network model are optimized using the backpropagation algorithm until the preset convergence condition is met, at which point the training ends and a trained next-order control strategy generation model is obtained.

[0011] In some embodiments, training an initial neural network model using the training sample set and optimizing the parameters of the initial neural network model using a backpropagation algorithm includes: During training, a pre-constructed loss function is used to evaluate the difference between the next-factor control policy output by the initial neural network model and the label of the next-factor control policy, and the weights and biases of the initial neural network model are adjusted according to the difference; the loss function is expressed as follows: ; in, This represents the value of the loss function. This represents the mean square error between the predicted value of the first control parameter and the first sample control parameter. This represents the mean square error between the predicted value of the second control parameter and the second sample control parameter. This represents the mean square error between the predicted value of the third control parameter and the third sample control parameter. This represents the absolute value of the difference between the predicted value of the control result score and the control result score in the next control strategy label. , , , These are the weighting coefficients.

[0012] In some embodiments, the method further includes: When the first control parameter in the unloading control strategy corresponding to the target package conflicts with the first control parameter in the unloading control strategy corresponding to other related unloading packages, the minimum or average value is taken as the compatibility parameter. Based on the compatibility parameter, the first control parameter corresponding to the target package and other related unloading packages that conflict with the target package is re-determined. At the same time, based on the change of the compatibility parameter compared to the original parameter, at least some parameters in the second control parameters corresponding to the target package and / or the other related unloading packages are adjusted.

[0013] In some embodiments, when the first control parameter in the unloading control strategy corresponding to the target package conflicts with the first control parameter in the unloading control strategy corresponding to other related unloading packages, the minimum or average value is taken as the compatibility parameter, including: Identify other related packages to be unloaded whose overlap ratio with the unloading control time period corresponding to the target package is greater than a first preset threshold, and obtain the first control parameters corresponding to the target package and the other related packages to be unloaded respectively; The maximum speed difference at each time point within the intersecting time period is calculated based on the movement speed in the first control parameters corresponding to the target package and the other related packages to be unloaded. When the maximum speed difference corresponding to all time points within the intersecting time period is less than the second preset threshold, the average speed corresponding to each time point within the intersecting time period is calculated to obtain the compatibility parameter; when the maximum speed difference corresponding to all time points within the intersecting time period is greater than or equal to the second preset threshold, the minimum speed corresponding to each time point within the intersecting time period is taken to obtain the compatibility parameter.

[0014] The beneficial effects of the method for dynamically adapting to different sizes of sorting compartments provided in the embodiments of this specification include at least the following: by inputting the weight data, volume characteristics, posture characteristics of the target package and the sorting compartment information matching the target package into a trained sorting control strategy generation model, a sorting control strategy for the target package is obtained. The control parameters can be dynamically adjusted according to the sorting compartment information and the weight data, volume characteristics, and posture characteristics of the target package, thereby dynamically adapting to the sorting process of different sizes of sorting compartments and achieving precise control in the sorting process.

[0015] Additional features will be set forth in part in the description which follows. They will become apparent to those skilled in the art upon consulting the following description and the accompanying drawings, or may be learned by the generation or operation of examples. The features of this specification can be realized and obtained through practice or by using various aspects of the methods, tools, and combinations illustrated in the following detailed examples. Attached Figure Description

[0016] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein: Figure 1 This is an exemplary flowchart of a method for dynamically adapting the size of the slots for component placement, as shown in some embodiments of this specification. Figure 2 These are exemplary structural diagrams of volume recognition networks and pose recognition networks shown in some embodiments of this specification; Figure 3 This is an exemplary structural diagram of a model generated according to some embodiments of the component control strategy shown in this specification; Figure 4 This is an exemplary flowchart of a method for dynamically adapting the size of slots for parts according to other embodiments of this specification. Detailed Implementation

[0017] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of this specification. For those skilled in the art, these drawings can be applied to other similar scenarios without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0018] It should be understood that the terms "system," "device," "unit," and / or "module" as used in this specification are a method of distinguishing different components, elements, parts, sections, or assemblies at different levels. However, if other terms can achieve the same purpose, they may be replaced by other expressions.

[0019] As indicated in this specification and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of expressly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0020] Flowcharts are used in this specification to illustrate the operations performed by the system according to embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the steps can be processed in reverse order or simultaneously. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0021] The method for dynamically adapting to different grid sizes provided in the embodiments of this specification will now be described in detail with reference to the accompanying drawings.

[0022] Figure 1 This is an exemplary flowchart of a method for dynamically adapting the size of the grid cutter according to some embodiments of this specification. In some embodiments, the method for dynamically adapting the size of the grid cutter can be executed by processing logic, which may include hardware (e.g., circuitry, dedicated logic, programmable logic, microcode, etc.), software (instructions running on a processing device to execute hardware simulations), and any combination thereof. In some embodiments, Figure 1 One or more operations in the flowchart of the method for dynamically adapting the size of grid slots can be implemented by a processing device and / or a terminal device. For example, the method for dynamically adapting the size of grid slots can be stored in a storage device in the form of a computer program and / or instructions, and can be invoked and / or executed by the processing device and / or the terminal device to implement the control process for dynamically adapting the size of grid slots as described below.

[0023] Reference Figure 1 The method for dynamically adapting to different grid sizes provided in this application embodiment may include the following steps S110~S140: Step S110: Obtain the weight data corresponding to the target package, as well as multiple package images collected from multiple different perspectives.

[0024] In this embodiment of the application, the weight data of the target package can be collected through the package feeding platform (the starting point of package sorting). At the same time, multiple (e.g., at least 3) image acquisition devices can be set above the package feeding platform. These multiple image acquisition devices can take pictures of the target package from different perspectives to obtain multiple package images corresponding to the target package. The purpose is to fully understand the appearance characteristics of the target package, so as to more accurately identify the size information of the target package in the subsequent process.

[0025] In some embodiments of this application, multiple image acquisition devices may be installed above the sorting loop to continuously capture dynamic images of the target packages during their movement. In some embodiments, to ensure the efficiency of image acquisition and the integrity of the data, the aforementioned multiple image acquisition devices may be equipped with autofocus and high frame rate shooting capabilities to cope with package flow at different speeds and angles, thereby effectively reducing data loss caused by environmental interference or equipment installation location limitations, and providing more reliable image data for subsequent processing.

[0026] Step S120: Obtain the volume and orientation features of the target package based on the multiple package images.

[0027] Since the volume and posture of a package can affect the accuracy of its landing in the corresponding slot during the unloading process (for example, packages of different volumes may require different unloading speeds (i.e., the driving speed at which the sorting platform drives the package toward the slot), while posture may cause the package to deviate or get stuck when it lands in the slot), in order to accurately control the relevant driving hardware during the unloading process, it is necessary to identify the volume and posture characteristics of the target package by using multiple package images collected from multiple different perspectives during the aforementioned process.

[0028] In some embodiments of this application, the aforementioned multiple package images can be processed using a neural network model to identify the volumetric and pose features of the target package. Specifically: First, the multiple package images can be preprocessed; then, the preprocessed multiple package images can be input into a trained volumetric recognition network and pose recognition network to extract the volumetric feature vector and pose feature vector corresponding to each package image; further, the volumetric feature vector and pose feature vector corresponding to the multiple package images can be integrated to obtain the comprehensive volumetric vector and comprehensive pose vector corresponding to the target package; finally, by mapping the comprehensive volumetric vector and the comprehensive pose vector, the volumetric and pose features corresponding to the target package are obtained.

[0029] In this embodiment of the application, the above preprocessing process may include denoising and scaling, wherein the denoising process includes removing background information from each package image except for the target package and the sorting platform it is located on.

[0030] As an example only, in some embodiments of this application, the Canny edge detection algorithm or the Sobel edge detection algorithm can be used to identify the target package outline corresponding to the target package in each package image, as well as the sorting platform outline intersecting with the target package outline; then, a mask image is generated based on the target package outline and the sorting platform outline (the mask image only retains the image information corresponding to the target package and its sorting platform); finally, the mask image is used to remove background information from the corresponding original package images in the multiple package images. It should be noted that, in the embodiments of this application, a mask image can be determined for each of the above multiple package images, and then the background information of the corresponding original package image can be removed based on the mask image, thereby obtaining the preprocessed image corresponding to each package image. In the embodiments of this application, by removing background information other than the target package and its sorting platform from each package image through the above process, interference factors in the subsequent volume feature and pose feature recognition process can be effectively reduced, thereby ensuring the accuracy of the volume feature and pose feature recognition results.

[0031] In this embodiment, scale unification refers to adjusting the resolution and size (affected by factors such as shooting distance and focal length) of each package image to be consistent, so that subsequent processing steps can perform feature extraction based on a unified standard. In this embodiment, this scale unification process can be implemented through various image processing techniques, such as using image scaling algorithms to adjust package images of different sizes to a preset standard resolution and size.

[0032] It should also be noted that, in the embodiments of this application, the above preprocessing may include, but is not limited to, the denoising and scaling processes described above. For example, in some embodiments, image enhancement processing may also be performed on the multiple package images to improve image clarity and detail.

[0033] After image preprocessing is completed, the preprocessed multiple package images can be input into the trained volume recognition network and pose recognition network to extract the volume feature vector and pose feature vector corresponding to each package image through the volume recognition network and pose recognition network.

[0034] Figure 2 This is an exemplary structural diagram of a volume recognition network and a pose recognition network according to some embodiments of this specification. (Refer to...) Figure 2In this embodiment, both the volume recognition network and the pose recognition network can be implemented based on a convolutional neural network architecture. The volume recognition network may include a first feature extraction layer, a first feature fusion layer, and a first feature mapping layer; the pose recognition network may include a second feature extraction layer, a second feature fusion layer, and a second feature mapping layer. In some embodiments of this application, the first and second feature extraction layers may each include convolutional layers and pooling layers, and the first and second feature mapping layers may be a fully connected layer.

[0035] Specifically, refer to Figure 2 In this embodiment, after inputting the preprocessed multiple package images into the trained volume recognition network, the multiple package images can be processed by a first feature extraction layer to extract the volume feature vector corresponding to each package image, resulting in multiple volume feature vectors. Further, the multiple volume feature vectors corresponding to the multiple package images can be integrated (e.g., by vector concatenation or fusion) by a first feature fusion layer to obtain a comprehensive volume vector corresponding to the target package. Finally, the comprehensive volume vector can be mapped by a first feature mapping layer to obtain the volume feature corresponding to the target package. In this embodiment, the volume feature may include the maximum and minimum cross-sectional dimensions corresponding to the target package.

[0036] Similarly, in this embodiment, after inputting the preprocessed multiple package images into the trained pose recognition network, the multiple package images can be processed by the second feature extraction layer to extract the pose feature vector corresponding to each package image, resulting in multiple pose feature vectors. Further, the multiple pose feature vectors corresponding to the multiple package images can be integrated (e.g., by vector concatenation or fusion) by the second feature fusion layer to obtain the comprehensive pose vector corresponding to the target package. Finally, the comprehensive pose vector can be mapped by the second feature mapping layer to obtain the pose feature corresponding to the target package. In this embodiment, the pose feature may include the center of gravity position of the target package and the relative positional relationship between the target package and its sorting platform.

[0037] In this embodiment, the volume recognition network and pose recognition network described above can be obtained through training. Specifically, multiple sample package images acquired from multiple different specified perspectives and their corresponding volume feature labels and pose feature labels can be combined to form training samples for training the volume recognition network and pose recognition network, respectively. As an example only, in some embodiments of this application, the maximum and minimum cross-sectional dimensions in the volume feature labels can refer to the length and width of the minimum bounding rectangle corresponding to the maximum and minimum cross-sections, respectively. The centroid position in the pose feature labels can be represented by its coordinate position in a three-dimensional coordinate system established by the sorting platform. The relative positional relationship between the target package and its sorting platform can be represented by the distance between the package outline boundary and the boundary of its sorting platform.

[0038] It is understood that, in the embodiments of this application, by constructing a large number of similar training samples to train the volume recognition network and the pose recognition network, the volume recognition network can acquire the ability to recognize the volume features of a package based on multiple package images, and the pose recognition network can acquire the ability to recognize the pose features of a package based on multiple package images.

[0039] It should be noted that during the training process of the aforementioned volume recognition network and pose recognition network, the parameters of the first feature extraction layer, first feature fusion layer, first feature mapping layer, second feature extraction layer, second feature fusion layer, and second feature mapping layer can be dynamically adjusted to optimize network performance and thereby improve the accuracy of volume and pose feature recognition results. Further details regarding the composition and training process of the aforementioned volume recognition network and pose recognition network can be considered prior art and will not be discussed in detail here.

[0040] Step S130: Match the lower compartment corresponding to the target package and obtain the corresponding lower compartment information.

[0041] In this embodiment, the target drop-off area can first be determined based on the barcode information corresponding to the target package (for example, the same target drop-off area can represent packages destined for the same region). Specifically, in this embodiment, the target drop-off area can include multiple compartments of different sizes, and packages sorted by different compartments can represent different delivery types or requirements (for example, large packages and small packages can be distinguished by the compartments).

[0042] In this embodiment, the closest size among the multiple compartments in the target unloading area can be determined as the unloading compartment based on the volume characteristics obtained in the above steps (e.g., the maximum cross-sectional dimension included in the aforementioned volume characteristics). Specifically, in this embodiment, the closest size compartment refers to the compartment whose corresponding minimum channel size is greater than the maximum cross-sectional dimension corresponding to the target package (both length and width directions must be satisfied), and whose length and width are closest to the maximum cross-sectional dimension.

[0043] After determining the unloading compartment using the above method, its corresponding unloading compartment information can be obtained. In this embodiment, the unloading compartment information is used to describe the distribution location and specifications of the unloading compartment. The distribution location can be represented in the form of coordinate positions (i.e., the coordinate position of the unloading compartment in a two-dimensional or three-dimensional coordinate system established by the sorting loop), and the compartment specifications include at least its corresponding minimum channel size (i.e., the size of the narrowest part of the compartment that can accommodate the passage of a package).

[0044] Step S140: Determine the unloading control strategy corresponding to the target package based on the unloading compartment information and the weight data, volume characteristics and posture characteristics of the target package.

[0045] In this embodiment of the application, the unloading control strategy includes a first control parameter for controlling the sorting loop to move in a circumferential direction, a second control parameter for controlling the sorting platform where the target package is located to drive the target package toward the unloading compartment, and a third control parameter for controlling the deviation correction system to correct the deviation of the target package.

[0046] In this embodiment, the information of the lower compartment, along with the weight, volume, and posture characteristics of the target package, can be used as input data to a trained lower compartment control strategy generation model. The model then outputs a lower compartment control strategy that matches the target package. (Refer to...) Figure 3 In this embodiment, the subordinate control strategy generation model may include a neural network model, which includes an input layer, an output layer, and multiple hidden layers. The input layer receives the input data, and the output layer outputs the first control parameter, the second control parameter, and the third control parameter. In this embodiment, the multiple hidden layers can be used for multi-level feature extraction and processing of the input data to model complex relationships. In this embodiment, the specific number of hidden layers and the number of neurons in each layer can be adjusted according to actual needs.

[0047] In some embodiments of this application, the first control parameter may include the movement speed, acceleration, and deceleration point of the sorting loop in the unloading stage; the second control parameter may include the driving force, driving speed, and driving direction of the sorting platform in the unloading stage; and the third control parameter may include the correction force and correction angle of the correction system.

[0048] Specifically, in this embodiment, the first control parameter in the unloading control strategy generated based on the above-mentioned unloading control strategy generation model can be used to control the movement speed (i.e., the circumferential movement speed of the sorting loop), acceleration, and deceleration point (the deceleration point can be regarded as the starting point of the unloading stage, which can be a coordinate point in the sorting loop located before the unloading slot) of the sorting loop. The second control parameter can be used to precisely control the driving force, driving speed, and driving direction (e.g., driving to the left or right, determined according to the positional relationship between the unloading slot and the sorting platform) of the sorting platform where the target package is located in the unloading stage, thereby ensuring that the target package can be accurately placed into the unloading slot at the appropriate time and location. In addition, in this embodiment, the third control parameter can be used to effectively avoid sorting errors that may be caused by package posture deviation by setting reasonable correction force and correction angle. In some embodiments, the third control parameter can be used to control the correction rollers set on the sorting platform. By coordinating the control of multiple correction rollers at different positions on the sorting platform, different correction angles can be achieved.

[0049] It should be noted that, in the embodiments of this application, the first, second, and third control parameters can all be determined based on the aforementioned unloading compartment information and the weight, volume, and posture characteristics of the target package. In other words, in the embodiments of this application, the specific information of the unloading compartment and the multidimensional characteristics of the target package can be comprehensively analyzed through the unloading control strategy generation model to dynamically generate control parameters adapted to the specifications of the unloading compartment. This method can not only improve the intelligence level of the sorting system but also significantly improve the accuracy and efficiency of the package sorting process. In some embodiments of this application, some of the parameters in the first, second, and third control parameters can be represented in the form of time-series data.

[0050] In addition, it should be noted that in some embodiments of this application, in order to further improve the adaptability of the above-mentioned unloading control strategy generation model, the output of the neural network model can be adaptively adjusted according to the relationship between different unloading slot specifications and package characteristics during actual training. For example, when facing a larger unloading slot, the requirements for package posture adjustment can be appropriately relaxed (i.e., the parameter range of the third control parameter can be appropriately relaxed); while for a smaller slot, the accuracy of each control parameter needs to be improved to ensure the accuracy of package sorting.

[0051] In this embodiment, the aforementioned subordinate control strategy generation model can be obtained through training. The training process of the subordinate control strategy generation model involved in this embodiment is briefly described below: First, a training sample set is constructed. In this embodiment, the training sample set includes multiple training sample data entries. Each training sample data entry may include the lower compartment information, weight data, volume characteristics, posture characteristics, and a lower control strategy label corresponding to the sample package. Specifically, in this embodiment, the lower control strategy label includes a first sample control parameter, a second sample control parameter, a third sample control parameter, and a corresponding control result score. The first sample control parameter, the second sample control parameter, the third sample control parameter, and the corresponding control result score can be obtained based on experimental testing (the control result score can be obtained through manual evaluation).

[0052] It should be noted that, in the embodiments of this application, a large amount of training sample data with different features (such as different lower part grid information, different weight data, different volume features and / or different posture features) can be collected, so that the lower part control strategy generation model can learn how to generate the optimal lower part control strategy according to the combination of different features.

[0053] In some embodiments of this application, the training samples may include positive samples and negative samples, wherein positive samples refer to sample data with higher control result scores, and negative samples refer to sample data with lower control result scores or sorting errors.

[0054] In some embodiments of this application, different control strategy labels can be configured for the same set of lower part grid information, weight data, volume characteristics, and posture characteristics. These different lower part control strategy labels can contain different control strategies and corresponding control result scores.

[0055] Considering the relatively complex process of collecting the aforementioned training sample data, in some embodiments of this application, more training sample data can be obtained through sample augmentation. For example, existing samples can be transformed or new sample data can be generated through simulation. Specifically, the weight data, volume characteristics, posture characteristics of the package, and the location and specification information corresponding to the lower compartment can be fine-tuned, or different combinations of lower compartment information and weight data can be simulated through a simulation system to generate diverse training samples. In addition, in some embodiments of this application, noisy data or extreme cases under boundary conditions can be introduced to improve the robustness and generalization ability of the model. It can be understood that, in the embodiments of this application, sample augmentation using the above methods can not only effectively expand the training sample set, but also help the aforementioned lower compartment control strategy generation model better adapt to the complex changes in actual application scenarios.

[0056] Furthermore, after obtaining the above training sample set, the initial neural network model can be trained using the training sample set, and the parameters of the initial neural network model can be optimized through the backpropagation algorithm until the preset convergence condition is met (e.g., the loss function reaches a preset threshold) and the training ends, thus obtaining the trained next-factor control strategy generation model.

[0057] During the training process of the aforementioned next-part control policy generation model, a pre-constructed loss function can be used to evaluate the difference between the next-part control policy output by the initial neural network model and the next-part control policy label, and the weights and biases of the initial neural network model can be adjusted according to the difference. Exemplarily, in some embodiments of this application, the loss function can be expressed as follows: ; in, This represents the value of the loss function. This represents the mean square error between the predicted value of the first control parameter in the next-factor control strategy output by the next-factor control strategy generation model and the first sample control parameter. This represents the mean square error between the predicted value of the second control parameter in the next-factor control strategy output by the next-factor control strategy generation model and the second sample control parameter. This represents the mean square error between the predicted value of the third control parameter in the next-factor control strategy output by the next-factor control strategy generation model and the third sample control parameter. This represents the absolute value of the difference between the predicted value of the control result score in the next-component control strategy output by the next-component control strategy generation model and the control result score in the label of the next-component control strategy. , , , These are the weighting coefficients.

[0058] In the embodiments of this application, the above-mentioned weighting coefficients , , , It can be adjusted according to actual needs. For example, in some scenarios, if the accuracy of the first control parameter is of greater concern, it can be appropriately increased. The value; if higher accuracy is required for the control result scoring, it can be increased. The proportion of.

[0059] Reference Figure 3 In this embodiment, the parameters of the subordinate control policy generation model can be optimized based on the loss calculation results obtained from the above-mentioned loss function. In this embodiment, the optimization algorithm used in the optimization process may include, but is not limited to, stochastic gradient descent, Adam optimization algorithm, etc. Further details regarding the structure and training process of the subordinate control policy generation model can be considered prior art and will not be elaborated upon in this specification.

[0060] After training is completed, the unloading compartment information (the unloading compartment information matching the target package) obtained in the above process, along with the weight data, volume characteristics, and posture characteristics of the target package, are input into the trained unloading control strategy generation model to obtain the unloading control strategy for the target package. It should be noted that, in this embodiment, the unloading control strategy can dynamically adjust control parameters based on the input feature information (such as the unloading compartment information and the weight data, volume characteristics, and posture characteristics of the target package), thereby dynamically adapting to the sorting process of different sized compartments and achieving precise control during the sorting process. As can be seen from the above, in this embodiment, the control parameters may include the movement speed, acceleration, and deceleration point of the sorting loop during the unloading stage, the driving force, driving speed, and driving direction of the sorting platform during the unloading stage, and the correction force and correction angle of the correction system.

[0061] In some embodiments of this application, sorting results can be collected in real time during practical applications. When a sorting error is detected, the corresponding unloading slot information, weight data, volume characteristics, posture characteristics, and unloading control strategy are used to form a negative sample (i.e., the corresponding control result score is 0). The unloading control strategy generation model is then trained to continuously improve the accuracy and adaptability of the unloading control strategy generation model.

[0062] Similarly, in some embodiments of this application, the control results of the package sorting process can be automatically scored based on machine vision technology. Then, the control result score, together with the corresponding unloading compartment information, weight data, volume characteristics, posture characteristics, and unloading control strategy, constitutes a new training sample, thereby continuously training and optimizing the unloading control strategy generation model.

[0063] In the embodiments of this application, each package to be sorted can be processed in the above manner to obtain its corresponding unloading control strategy. However, when there are a large number of packages to be sorted at the same time, conflicts may occur (mainly conflicts in the first control parameters of the unloading control strategy). To address this situation, in some embodiments of this application, when the first control parameter in the unloading control strategy corresponding to the target package conflicts with the first control parameter in the unloading control strategy corresponding to other related unloading packages, the minimum or average value can be taken as a compatibility parameter. Based on the compatibility parameter, the first control parameter commonly corresponding to the target package and other related unloading packages that conflict with the target package can be re-determined. At the same time, at least some parameters in the second control parameters corresponding to the target package and / or the other related unloading packages can be adjusted based on the change of the compatibility parameter compared to the original parameter.

[0064] Figure 4 This is an exemplary flowchart of a method for dynamically adapting the size of slots for parts according to other embodiments of this specification. (Refer to...) Figure 4 In some embodiments of this application, the handling of the above-mentioned conflict situation may include: Step S150: Determine other related packages to be unloaded whose overlap ratio with the unloading control time period corresponding to the target package is greater than a first preset threshold, and obtain the first control parameters corresponding to the target package and the other related packages to be unloaded respectively.

[0065] In this embodiment, the unloading control time period corresponding to the target package can be understood as the time period between the start and end times of the system's control over the unloading process of the target package. In this embodiment, when the overlap ratio between the unloading control time periods corresponding to other packages awaiting unloading and the unloading control time period corresponding to the target package is greater than a first preset threshold (e.g., 50% or other values), it can be considered a conflict. In this case, a compatibility scheme needs to be further determined based on the relationship between the first control parameters corresponding to other related packages awaiting unloading and the first control parameters corresponding to the target package.

[0066] Step S160: Calculate the maximum speed difference at each time point within the intersecting time period based on the movement speed in the first control parameters corresponding to the target package and the other related packages to be unloaded.

[0067] Specifically, in this embodiment of the application, the maximum speed difference at each point in time within the intersecting time period can be calculated based on the speed of movement in the first control parameters corresponding to the target package and the other related packages to be unloaded. The larger the maximum speed difference, the more serious the conflict; conversely, the smaller the maximum speed difference, the less serious the conflict.

[0068] Step S170: When the maximum speed difference corresponding to all time points within the intersecting time period is less than the second preset threshold, calculate the average speed corresponding to each time point within the intersecting time period to obtain the compatibility parameter; when the maximum speed difference corresponding to all time points within the intersecting time period is greater than or equal to the second preset threshold, take the minimum speed corresponding to each time point within the intersecting time period to obtain the compatibility parameter.

[0069] In this embodiment, the severity of the conflict can be determined by setting the aforementioned second preset threshold, and the movement speed of the sorting loop can be adjusted accordingly to achieve a compatible unloading effect, thereby minimizing interference with each other's sorting processes. Specifically, in some embodiments of this application, if the aforementioned maximum movement speed difference is less than the second preset threshold, it indicates a minor conflict, and the average speed corresponding to each time point within the intersecting time period can be calculated as a compatibility parameter; conversely, if the aforementioned maximum movement speed difference is greater than or equal to the second preset threshold, it indicates a more severe conflict, and the minimum speed corresponding to each time point within the intersecting time period can be selected as a compatibility parameter. In this embodiment, the specific value of the aforementioned second preset threshold can be set according to actual needs, and this specification does not specifically limit it.

[0070] In this embodiment, considering that using the aforementioned compatible parameters for sorting loop control may also affect the control effect of the second control parameter in the unloading control strategy, to address this issue, in some embodiments of this application, at least some parameters in the second control parameter corresponding to the target package and / or other related unloading packages can be adjusted based on the magnitude of change of the compatible parameters compared to the original parameters.

[0071] In summary, the beneficial effects that the embodiments of this specification may bring include, but are not limited to: (1) In the method of dynamically adapting to different sizes of slots for unloading provided in some embodiments of this specification, by inputting the weight data, volume characteristics, posture characteristics of the target package and the unloading slot information that matches the target package into the trained unloading control strategy generation model, an unloading control strategy for the target package is obtained. The control parameters can be dynamically adjusted according to the unloading slot information and the weight data, volume characteristics, and posture characteristics of the target package, thereby dynamically adapting to the sorting process of different sizes of slots and achieving precise control in the sorting process; (2) In the method of dynamically adapting to different sizes of slots for unloading provided in some embodiments of this specification In this method, when the first control parameter in the unloading control strategy corresponding to the target package conflicts with the first control parameter in the unloading control strategy corresponding to other related unloading packages, the minimum or average value is taken as the compatibility parameter. Based on the compatibility parameter, the first control parameter corresponding to the target package and other related unloading packages is re-determined. At the same time, based on the change of the compatibility parameter compared to the original parameter, at least some parameters in the second control parameters corresponding to the target package and / or other related unloading packages are adjusted. This can effectively solve the problem of control parameter conflicts when multiple packages are unloaded at the same time, thereby ensuring sorting accuracy while minimizing equipment downtime or incorrect sorting caused by parameter conflicts.

[0072] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects may be any one or a combination of the above, or any other possible beneficial effects.

[0073] The basic concepts have been described above. Obviously, for those skilled in the art, the detailed disclosure above is merely illustrative and does not constitute a limitation of this specification. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and corrections to this specification. Such modifications, improvements, and corrections are suggested in this specification and therefore remain within the spirit and scope of the exemplary embodiments described herein.

[0074] Furthermore, this specification uses specific terms to describe embodiments thereof. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "an embodiment," "one embodiment," or "an alternative embodiment" in different locations throughout this specification do not necessarily refer to the same embodiment. Moreover, certain features, structures, or characteristics in one or more embodiments of this specification can be appropriately combined.

[0075] Furthermore, those skilled in the art will understand that various aspects of this specification can be described and illustrated in several patentable ways or situations, including any new and useful combination of processes, machines, products, or substances, or any new and useful improvements thereof. Accordingly, various aspects of this specification can be implemented entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. All of the above hardware or software may be referred to as a “data block,” “module,” “engine,” “unit,” “component,” or “system.” Furthermore, various aspects of this specification may be represented as a computer product located on one or more computer-readable media, including computer-readable program code.

[0076] Computer storage media may contain a propagated data signal containing computer program code, for example, on baseband or as part of a carrier wave. This propagated signal may take various forms, including electromagnetic, optical, and suitable combinations thereof. Computer storage media can be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, apparatus, or device to enable communication, propagation, or transmission of a program for use. The program code located on the computer storage medium can be propagated through any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of the above media.

[0077] The computer program code required for the operation of each part of this manual can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc.; conventional procedural programming languages ​​such as C, Visual Basic, Fortran2003, Perl, COBOL2002, PHP, ABAP; dynamic programming languages ​​such as Python, Ruby, and Groovy; or other programming languages. This program code can run entirely on the user's computer, or as a standalone software package on the user's computer, or partially on the user's computer and partially on a remote computer, or entirely on a remote computer or processing device. In the latter case, the remote computer can be connected to the user's computer through any network, such as a local area network (LAN) or wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as Software as a Service (SaaS).

[0078] Furthermore, unless expressly stated in the claims, the order of processing elements and sequences, the use of numbers and letters, or other names described in this specification are not intended to limit the order of the processes and methods described herein. Although various examples have been discussed in the foregoing disclosure of some embodiments of the invention that are currently considered useful, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments; rather, the claims are intended to cover all modifications and equivalent combinations that conform to the spirit and scope of the embodiments described herein. For example, while the system components described above can be implemented by hardware devices, they can also be implemented solely by software solutions, such as installing the described system on existing processing devices or mobile devices.

[0079] Similarly, it should be noted that, in order to simplify the description disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments in this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0080] In some embodiments, numbers describing the quantity of components and attributes are used. It should be understood that such numbers used in the description of embodiments are modified in some examples with the terms "approximately," "approximately," or "generally." Unless otherwise stated, "approximately," "approximately," or "generally" indicates that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may be changed depending on the characteristics required by individual embodiments. In some embodiments, numerical parameters should take into account specified significant digits and employ a general method of digit reservation. Although the numerical ranges and parameters used to confirm their breadth of range in some embodiments of this specification are approximate values, in specific embodiments, such values ​​are set as precisely as feasible.

[0081] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A method of dynamically adapting to a size pocket unloading, characterized in that, The method comprises the following steps: acquiring weight data corresponding to a target package and a plurality of package images of the target package collected from different perspectives; acquiring volume features and posture features of the target package based on the plurality of package images; matching a corresponding unloading compartment of the target package and acquiring corresponding unloading compartment information, wherein the unloading compartment information is used to describe the distribution position and compartment specifications corresponding to the unloading compartment; determining an unloading control strategy corresponding to the target package according to the unloading compartment information and the weight data, volume features and posture features corresponding to the target package, wherein the unloading control strategy comprises a first control parameter for controlling the ring line to move in a ring direction, a second control parameter for controlling the target package to move towards the unloading compartment by driving the target package on the sorting platform, and a third control parameter for controlling the target package to be corrected by the correction system.

2. The method of claim 1, wherein, The method comprises the following steps: preprocessing the plurality of package images, wherein the preprocessing at least comprises denoising processing and scale unification, and the denoising processing comprises removing background information in each of the package images except for the target package and the sorting platform on which the target package is located; inputting the preprocessed plurality of package images into a trained volume recognition network and a posture recognition network to extract a volume feature vector and a posture feature vector corresponding to each of the package images through the volume recognition network and the posture recognition network; integrating the volume feature vectors and the posture feature vectors corresponding to the plurality of package images to obtain a comprehensive volume vector and a comprehensive posture vector corresponding to the target package; mapping the comprehensive volume vector and the comprehensive posture vector to obtain volume features and posture features corresponding to the target package, wherein the volume features comprise maximum and minimum cross-sectional dimensions corresponding to the target package, and the posture features comprise a center of gravity position of the target package and a relative position relationship between the target package and the sorting platform on which the target package is located.

3. The method of claim 2, wherein, The method comprises the following steps: identifying a target package contour corresponding to the target package and a sorting platform contour intersecting the target package contour in each of the package images through an edge detection algorithm, wherein the edge detection algorithm comprises a Canny edge detection algorithm or a Sobel edge detection algorithm; generating a mask image according to the target package contour and the sorting platform contour, wherein only image information corresponding to the target package and the sorting platform on which the target package is located is retained in the mask image; removing background information from the corresponding original package images in the plurality of package images by using the mask image.

4. The method of claim 2, wherein, The method comprises the following steps: determining a target unloading area according to the barcode information corresponding to the target package, wherein the target unloading area comprises a plurality of compartments of different specifications; Determine a size closest pocket as a lower pocket from a plurality of pockets contained in the target lower part area according to the volume characteristics, wherein the minimum channel size corresponding to the lower pocket is greater than the maximum cross-sectional size corresponding to the target package.

5. The method of claim 1, wherein, The method further comprises: Input the lower pocket information and the weight data, volume characteristics and attitude characteristics corresponding to the target package as input data into the trained lower control strategy generation model to output the lower control strategy matched with the target package through the lower control strategy generation model. The lower control strategy generation model comprises a neural network model, and the neural network model comprises an input layer, an output layer and a plurality of hidden layers, wherein the input layer is used to receive the input data, and the output layer is used to output the first control parameter, the second control parameter and the third control parameter.

6. The method of adapting the size of the aperture for the piece of claim 5, wherein, The first control parameter comprises the motion speed, acceleration and deceleration point of the sorting loop in the lower part stage; the second control parameter comprises the driving force, driving speed and driving direction of the sorting platform in the lower part stage; and the third control parameter comprises the deviation correction force and deviation correction angle of the deviation correction system.

7. The method of claim 5, wherein, The lower control strategy generation model is trained based on the following manner: A training sample set is constructed, and the training sample set comprises a plurality of training sample data. Each piece of training sample data comprises lower pocket information, weight data, volume characteristics, attitude characteristics corresponding to a sample package and a lower control strategy label corresponding to the sample package. The lower control strategy label comprises a first sample control parameter, a second sample control parameter, a third sample control parameter and a corresponding control result score. The first sample control parameter, the second sample control parameter, the third sample control parameter and the corresponding control result score are obtained based on experimental tests. The initial neural network model is trained by using the training sample set, and the parameters of the initial neural network model are optimized by using a back propagation algorithm until a preset convergence condition is met, so as to obtain the trained lower control strategy generation model.

8. The method of claim 7, wherein, The initial neural network model is trained by using the training sample set, and the parameters of the initial neural network model are optimized by using a back propagation algorithm until a preset convergence condition is met, so as to obtain the trained lower control strategy generation model. In the training process, a loss function is used to evaluate the difference between the lower control strategy output by the initial neural network model and the lower control strategy label, and the weights and biases of the initial neural network model are adjusted according to the difference. The loss function is expressed as follows: ； wherein, represents a loss function value, represents a mean square error between the predicted value of the first control parameter and the first sample control parameter, represents a mean square error between the predicted value of the second control parameter and the second sample control parameter, represents a mean square error between the predicted value of the third control parameter and the third sample control parameter, represents an absolute value of a difference between the predicted value of the control result score and the control result score in the lower garment control strategy label, , , , is a weight coefficient.

9. The method of adapting the size of the opening for the fitting according to any one of claims 1 to 8, characterized in that, The method further comprises: When the first control parameter in the target package corresponding unpacking control strategy conflicts with the first control parameter in the unpacking control strategy corresponding to other related to-be-unpacked packages, the minimum value or the average value is taken as the compatible parameter, and the first control parameter corresponding to the target package and other related to-be-unpacked packages is re-determined based on the compatible parameter, and at least part of the second control parameters corresponding to the target package and / or the other related to-be-unpacked packages are adjusted based on the change amplitude of the compatible parameter compared with the original parameter.

10. The method of adapting the size of the aperture for the installation of the component according to claim 9, characterized in that, The first control parameter in the target package corresponding unpacking control strategy conflicts with the first control parameter in the unpacking control strategy corresponding to other related to-be-unpacked packages, and the minimum value or the average value is taken as the compatible parameter, including: determining other related to-be-unpacked packages corresponding to the target package corresponding unpacking control time period, and obtaining the first control parameter corresponding to the target package and the other related to-be-unpacked packages respectively; based on the motion speed in the first control parameter corresponding to the target package and the other related to-be-unpacked packages respectively, calculating the maximum motion speed difference value corresponding to each time point in the intersection time period; when the maximum motion speed difference value corresponding to all time points in the intersection time period is less than a second preset threshold, calculating the average speed corresponding to each time point in the intersection time period to obtain the compatible parameter; when the maximum motion speed difference value corresponding to all time points in the intersection time period is greater than or equal to the second preset threshold, taking the minimum speed corresponding to each time point in the intersection time period to obtain the compatible parameter.

Images