Robot scheduling method and device, equipment and medium

Abstract

The embodiment of the invention discloses a robot scheduling method and device, equipment and a medium, and relates to the technical field of artificial intelligence. The method comprises the following steps: acquiring current patrol data of a current patrol period from a main robot in a robot scheduling system; wherein the current patrol data is determined through a historical customer number obtained by the main robot in a previous patrol period and a preset customer number threshold value; according to the current patrol data, obtaining a customer face image of each customer to be served in the corresponding service area, inputting the customer face image into a trained face emotion recognition model, and outputting a candidate face emotion of the customer to be served and a corresponding candidate emotion probability; and determining a target facial emotion of the to-be-served customer from the candidate facial emotion according to the candidate emotion probability, and determining a target service customer and a corresponding target service strategy from the to-be-served customer according to the target facial emotion. According to the scheme, the accuracy and timeliness of robot scheduling are improved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of artificial intelligence technology, and in particular to a robot scheduling method, device, equipment and medium. Background Art

[0002] Driven by the wave of financial technology, the banking industry is undergoing unprecedented changes. Intelligent service innovation has become a key way to enhance bank competitiveness and optimize customer experience. Bank outlets are gradually transforming from traditional service models to intelligent and personalized ones. This transformation requires not only banks to innovate in technology applications, but also to achieve comprehensive upgrades in multiple dimensions such as service processes, product design and customer interaction.

[0003] In the existing technology, in order to reduce the daily workload of practitioners and provide customers with a more professional and considerate business service experience, an intelligent robot scheduling method is proposed, in which robots actively provide services to customers. Therefore, it is crucial to improve the accuracy and timeliness of robot scheduling in order to bring better service experience to customers. Summary of the invention

[0004] The present invention provides a robot scheduling method, device, equipment and medium to improve the accuracy and timeliness of robot scheduling and bring better service experience to customers.

[0005] According to one aspect of the present invention, a robot scheduling method is provided, which is applied to a slave robot in a robot scheduling system, and the method comprises:

[0006] Acquire current patrol data of a current patrol cycle from a master robot in the robot dispatching system; wherein the current patrol data is determined by a historical number of customers acquired by the master robot in a previous patrol cycle and a preset customer number threshold;

[0007] According to the current patrol data, obtain the customer facial images of each customer to be served in the corresponding service area, input the customer facial images into the trained facial emotion recognition model, and output the candidate facial emotions of the customer to be served and the corresponding candidate emotion probabilities;

[0008] Determine the target facial emotion of the customer to be served from the candidate facial emotions according to the candidate emotion probabilities, and determine the target service customer and the corresponding target service strategy from the customers to be served according to the target facial emotion;

[0009] The facial emotion recognition model is trained based on sample facial images of sample training objects and corresponding sample actual labels.

[0010] According to another aspect of the present invention, there is provided a robot scheduling device, which is configured for a slave robot in a robot scheduling system, and the device comprises:

[0011] A patrol data acquisition module, used to acquire current patrol data of a current patrol cycle from a master robot in the robot scheduling system; wherein the current patrol data is determined by the number of historical customers acquired by the master robot in a previous patrol cycle and a preset customer number threshold;

[0012] A model output module is used to obtain customer facial images of each customer to be served in the corresponding service area according to the current patrol data, and input the customer facial images into the trained facial emotion recognition model to output candidate facial emotions of the customer to be served and corresponding candidate emotion probabilities;

[0013] A service strategy determination module, used to determine the target facial emotion of the customer to be served from the candidate facial emotions according to the candidate emotion probability, and determine the target service customer and the corresponding target service strategy from the customers to be served according to the target facial emotion;

[0014] The facial emotion recognition model is trained based on sample facial images of sample training objects and corresponding sample actual labels.

[0015] According to another aspect of the present invention, there is provided an electronic device, comprising:

[0016] one or more processors;

[0017] A memory for storing one or more programs;

[0018] When one or more programs are executed by one or more processors, the one or more processors can execute any one of the robot scheduling methods provided by the embodiments of the present invention.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement any robot scheduling method provided by an embodiment of the present invention when executed.

[0020] An embodiment of the present invention provides a robot scheduling solution, which obtains current patrol data of a current patrol cycle from a master robot in a robot scheduling system; wherein the current patrol data is determined by the number of historical customers obtained by the master robot in the previous patrol cycle and a preset customer number threshold; based on the current patrol data, customer facial images of each customer to be served in a corresponding service area are obtained, and the customer facial images are input into a trained facial emotion recognition model, and candidate facial emotions of the customer to be served and corresponding candidate emotion probabilities are output; based on the candidate emotion probabilities, target facial emotions of the customer to be served are determined from the candidate facial emotions, and based on the target facial emotions, target service customers and corresponding target service strategies are determined from the customers to be served; wherein the facial emotion recognition model is trained based on sample facial images of sample training objects and corresponding sample actual labels. In the above scheme, the robot scheduling system includes a master robot and a slave robot. The slave robot patrols and obtains customer facial images in the corresponding service area through the current patrol data obtained from the master robot. Subsequently, the target service customers that the slave robot needs to serve in the service area are determined based on the customer facial images and facial emotion recognition models, thereby improving the accuracy of robot scheduling; and, the current patrol data in the current patrol cycle in the embodiment of the present invention is predicted by the master robot based on the historical number of customers in the previous patrol cycle and a preset customer number threshold. The slave robot obtains the customer facial images of each customer to be served in the corresponding service area based on the current patrol data, thereby realizing dynamic adjustment of the current patrol data of the slave robot in the current patrol cycle, improving the effectiveness of obtaining customer facial images, thereby improving the accuracy of subsequent determination of target service customers based on customer facial images and facial emotion recognition models, and improving the timeliness of robot scheduling.

[0021] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 is a flow chart of a robot scheduling method provided by Embodiment 1 of the present invention;

[0024] Figure 2 is a flow chart of a robot scheduling method provided by Embodiment 2 of the present invention;

[0025] Figure 3 is a system module diagram of a robot scheduling system provided by Embodiment 3 of the present invention;

[0026] Figure 4 is a structural schematic diagram of a robot scheduling device provided by a fourth embodiment of the present invention;

[0027] Figure 5 It is a structural schematic diagram of an electronic device for implementing a robot scheduling method provided in Embodiment 5 of the present invention. DETAILED DESCRIPTION

[0028] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0029] Embodiment 1

[0030] Figure 1 This is a flow chart of a robot scheduling method provided in Embodiment 1 of the present invention. This embodiment can be applied to the situation of scheduling robots in financial institutions. The method can be executed by a robot scheduling device, which can be implemented in software and / or hardware and can be configured in an electronic device that carries the robot scheduling function.

[0031] See also Figure 1 The robot scheduling method shown is applied to a slave robot in a robot scheduling system, and includes:

[0032] S110. Acquire current patrol data of a current patrol cycle from a master robot in the robot scheduling system.

[0033] Among them, the robot dispatching system can be used to dispatch robots. The master robot refers to the robot at the entrance of the financial institution. The master robot can guide customers to the corresponding service area according to the needs of customers entering the financial institution. The slave robot refers to the robot inside the financial institution. The slave robot can provide services for customers in the corresponding service area.

[0034] It should be noted that the robot dispatching system includes a master robot and multiple slave robots, and the master robot and the slave robots are connected in communication. One service area can correspond to one slave robot, that is, one slave robot can patrol the service area it is responsible for.

[0035] The current patrol cycle refers to the patrol cycle of the slave robot at the current moment. The current patrol data can be used to indicate the patrol data of the slave robot in the current patrol cycle. The current patrol data is determined by the number of historical customers obtained by the master robot in the previous patrol cycle and the preset customer number threshold.

[0036] The previous patrol cycle refers to the previous patrol cycle of the slave robot. For example, if the current patrol cycle is 9:00-10:00, the previous patrol cycle is 8:00-9:00. The number of historical customers refers to the number of customers entering the financial institution acquired by the master robot in the previous patrol cycle. The embodiment of the present invention does not impose any limitation on the size of the preset customer number threshold, which can be set by technicians based on experience or repeatedly determined through a large number of experiments.

[0037] It should be noted that the embodiment of the present invention does not impose any limitation on the length of the patrol cycle, which can be set by technicians based on experience. For example, a patrol cycle can be one hour.

[0038] In an optional embodiment, the current patrol data is determined in the following manner: comparing the historical customer quantity with a preset customer quantity threshold; and determining the current patrol data of the slave robot in the current patrol cycle based on the comparison result.

[0039] It can be understood that by comparing the historical number of customers with the preset customer number threshold to determine the current patrol data, real-time dynamic adjustment of the current patrol data is achieved, thereby improving the accuracy and reliability of the determined current patrol data.

[0040] In an optional embodiment, if the current patrol data includes the current patrol number and the current patrol time, the current patrol data of the slave robot in the current patrol cycle is determined based on the comparison result, including: if the comparison result is that the historical customer number is greater than or equal to the preset customer number threshold, then the basic patrol number of the slave robot in the current patrol cycle is obtained; based on the basic patrol number, the historical customer number, the preset customer number threshold and the preset frequency increase coefficient, the current patrol number of the slave robot in the current patrol cycle is determined, and the current patrol time is determined based on the current patrol number and the duration of the current patrol cycle; wherein the current patrol number is an integer, and the current patrol number is greater than or equal to the basic patrol number.

[0041] The basic patrol times refer to the preset minimum patrol times of the slave robot in any patrol cycle. The embodiment of the present invention does not impose any limitation on the size of the preset frequency increase coefficient, which can be set by technicians based on experience or repeatedly determined through a large number of experiments.

[0042] The current patrol number refers to the number of patrols of the slave robot in the current patrol cycle. The current patrol time refers to the patrol time point of the slave robot in the current patrol cycle.

[0043] For example, the current patrol times can be determined by the following formula:

[0044]

[0045] Where P represents the current patrol times; P 频率 Indicates the preset frequency increase coefficient; P 客户 represents the current number of customers; P' represents the preset customer number threshold.

[0046] Furthermore, according to the current patrol times, the duration of the current patrol cycle is equally divided to obtain the current patrol times and the current patrol time.

[0047] It can be understood that by determining the current patrol number of the slave robot in the current patrol cycle according to the basic patrol number, the historical customer number, the preset customer number threshold and the preset frequency increase coefficient when the comparison result is that the historical customer number is greater than or equal to the preset customer number threshold, the accuracy of the determined current patrol number is improved; at the same time, the present invention improves the applicability of the current patrol number by limiting the current patrol number to an integer and the current patrol number is greater than or equal to the basic patrol number, thereby avoiding the situation where the determined current patrol number cannot be used.

[0048] In another optional embodiment, if the comparison result is that the number of historical customers is less than the preset customer number threshold, the basic patrol number is used as the current patrol number in the current patrol cycle, and the current patrol time is determined based on the current patrol number and the duration of the current patrol cycle.

[0049] The setting of the patrol cycle in the embodiment of the present invention can be adjusted dynamically. For example, if the preset patrol cycle is 60 minutes, and the number of historical customers collected by the main robot in the previous patrol cycle is greater than the preset cycle adjustment threshold, the patrol cycle can be shortened according to the preset cycle adjustment step length to obtain the current patrol cycle. The preset cycle adjustment step length refers to the length of time for adjusting the patrol cycle once. For example, if the preset cycle adjustment step length can be 15 minutes, the current patrol cycle obtained after shortening is 45 minutes. The embodiment of the present invention does not impose any restrictions on the size of the preset cycle adjustment threshold and the preset cycle adjustment step length. They can be set by technicians based on experience or needs, or determined repeatedly through a large number of experiments.

[0050] S120, obtaining customer facial images of each customer to be served in the corresponding service area according to the current patrol data, and inputting the customer facial images into the trained facial emotion recognition model, and outputting candidate facial emotions of the customer to be served and the corresponding candidate emotion probabilities.

[0051] Among them, the service area refers to the waiting area corresponding to a certain business service provided to customers, that is, customers can wait for business processing in the service area that provides corresponding business services according to business needs.

[0052] The customer facial image refers to the facial image of the customer to be served in the service area. The embodiment of the present invention does not impose any limitation on the method of obtaining the customer facial image, which can be set by the technician based on experience. Exemplarily, the slave robot can directly obtain the customer facial image of the customer to be served in the service area it is responsible for, or the slave robot can obtain the customer image of the customer to be served in the service area it is responsible for, and use the target detection algorithm to process the customer image to obtain the customer facial image. The customer image refers to the image of the customer to be served.

[0053] Among them, customers waiting to be served refer to customers waiting in the service area.

[0054] The facial emotion recognition model can be used to recognize the facial emotions of the customer to be served. Exemplarily, the facial emotion recognition model is trained based on the sample facial images of the sample training object and the corresponding sample actual labels.

[0055] The sample training object refers to a customer who can provide training samples for the facial emotion recognition model. The sample facial image refers to the facial image of the sample customer. The sample actual label refers to the real emotion expressed by the sample customer in the sample facial image.

[0056] The candidate facial emotion refers to the possible emotion of the customer to be served predicted by the facial emotion recognition model, and the candidate emotion probability refers to the predicted probability corresponding to the candidate facial emotion.

[0057] S130 , determining a target facial emotion of the customer to be served from the candidate facial emotions according to the candidate emotion probabilities, and determining a target service customer and a corresponding target service strategy from the customers to be served according to the target facial emotion.

[0058] The target facial emotion refers to the candidate facial emotion with the highest probability among the candidate emotions. The target service customer refers to the customer to be served who needs to be served by the robot. The target service strategy refers to the strategy for providing services to the target service customer from the robot. The embodiment of the present invention does not impose any limitation on the target service strategy, which can be set by a technician based on experience.

[0059] In an optional embodiment, target service customers and corresponding target service strategies are determined from customers to be served based on target facial emotions, including: taking customers to be served whose target facial emotions are negative facial emotions as target service customers; determining the business type of the target service customer, and determining the target service strategy for the target service customer based on the business type and the negative facial emotion.

[0060] Wherein, negative facial emotions refer to negative facial emotions. For example, negative facial emotions may include anxiety and anger.

[0061] The business type refers to the type of business that needs to be processed for the target service customers.

[0062] For example, the slave robot can obtain the business needs of the target service customer and determine the business type of the target service customer based on the business needs; or the slave robot can send the customer number of the target service customer to the master robot, and the master robot can query and obtain the business type of the target service customer based on the customer number and feed the business type back to the slave robot. The customer number can be used to represent the identity of the customer.

[0063] It can be understood that by taking customers with negative facial emotions as target service customers and determining target service strategies based on business types, screening of customers to be served is achieved to provide customers with better services and improve customer experience.

[0064] An embodiment of the present invention provides a robot scheduling solution, which obtains current patrol data of a current patrol cycle from a master robot in a robot scheduling system; wherein the current patrol data is determined by the number of historical customers obtained by the master robot in the previous patrol cycle and a preset customer number threshold; based on the current patrol data, customer facial images of each customer to be served in a corresponding service area are obtained, and the customer facial images are input into a trained facial emotion recognition model, and candidate facial emotions of the customer to be served and corresponding candidate emotion probabilities are output; based on the candidate emotion probabilities, target facial emotions of the customer to be served are determined from the candidate facial emotions, and based on the target facial emotions, target service customers and corresponding target service strategies are determined from the customers to be served; wherein the facial emotion recognition model is trained based on sample facial images of sample training objects and corresponding sample actual labels. In the above scheme, the robot scheduling system includes a master robot and a slave robot. The slave robot patrols and obtains customer facial images in the corresponding service area through the current patrol data obtained from the master robot. Subsequently, the target service customers that the slave robot needs to serve in the service area are determined based on the customer facial images and facial emotion recognition models, thereby improving the accuracy of robot scheduling; and, the current patrol data in the current patrol cycle in the embodiment of the present invention is predicted by the master robot based on the historical number of customers in the previous patrol cycle and a preset customer number threshold. The slave robot obtains the customer facial images of each customer to be served in the corresponding service area based on the current patrol data, thereby realizing dynamic adjustment of the current patrol data of the slave robot in the current patrol cycle, improving the effectiveness of obtaining customer facial images, thereby improving the accuracy of subsequent determination of target service customers based on customer facial images and facial emotion recognition models, and improving the timeliness of robot scheduling.

[0065] On the basis of the above embodiment, the current patrol data of any slave robot in the embodiment of the present invention may also be determined by the master robot in the previous patrol cycle according to the number of historical regional customers directed to the service area where the slave robot is located and the threshold of the number of preset regional customers corresponding to the slave robot. Specifically, it may be determined based on the method for determining the current patrol number and the current patrol time in the embodiment of the present invention.

[0066] The number of historical regional customers refers to the number of customers in a service area in the previous patrol cycle. The embodiment of the present invention does not impose any limitation on the threshold of the number of customers in the preset area, which can be set by technicians based on experience or repeatedly determined through a large number of experiments.

[0067] Embodiment 2

[0068] Figure 2It is a flow chart of a robot scheduling method provided by Example 2 of the present invention. Based on the above embodiments, this embodiment further adds "inputting the sample facial image into the constructed facial emotion recognition model to obtain a candidate feature vector graph, and dividing the candidate feature vector graph into regions according to the height and width of the candidate feature vector graph to obtain a candidate feature vector subgraph; wherein the facial emotion recognition model includes a facial feature generation network, a two-layer attention mechanism and a fully connected layer; according to the candidate feature vector subgraph, the preset graph query weight, the preset graph key weight and the preset graph value weight, respectively determining the graph query data, the graph key data and the graph value data, and determining the target feature vector subgraph from the candidate feature vector subgraph according to the graph query data and the graph key data; determining the sample feature vector graph according to the target feature vector subgraph, the graph key data and the graph value data, and determining the sample prediction result of the sample training object in the sample facial image according to the sample feature vector graph; determining the model loss value of the facial emotion recognition model according to the sample prediction result of the sample training object and the corresponding sample actual label, and training the facial emotion recognition model according to the model loss value" to improve the training mechanism of the facial emotion recognition model. It should be noted that, for parts not described in detail in the embodiment of the present invention, reference may be made to descriptions in other embodiments.

[0069] See also Figure 2 The robot scheduling method shown is applied to a slave robot in a robot scheduling system, and includes:

[0070] S210, inputting the sample facial image into the constructed facial emotion recognition model to obtain a candidate feature vector map, and dividing the candidate feature vector map into regions according to the height and width of the candidate feature vector map to obtain candidate feature vector sub-maps.

[0071] Among them, the facial emotion recognition model includes a facial feature generation network, a two-layer attention mechanism and a fully connected layer.

[0072] Among them, the candidate feature vector map refers to a vector feature map obtained by extracting features and converting vectors of sample facial images. Exemplarily, the sample facial image is input into the facial feature generation network in the facial emotion recognition model to obtain a candidate feature vector map. The facial feature generation network can be used to extract key facial features from the sample facial image and perform vector conversion to obtain a candidate feature vector map. The embodiment of the present invention does not impose any limitation on the network structure adopted by the facial feature generation network, which can be set by technicians based on experience. Exemplarily, a lightweight network can be used as the basis, that is, a combination of residual blocks and non-residual blocks to obtain a facial feature generation network.

[0073] It should be noted that when the sample facial image is input into the facial feature generation network, all network structures in the facial feature generation network participate in the processing, and a 256-dimensional candidate feature vector map is obtained; or, when the sample facial image is input into the facial feature generation network, the network structures except the last three layers in the facial feature generation network participate in the processing, and a 14×14×128 candidate feature vector map is obtained. The advantage is that it avoids the feature loss caused by the last three layers when processing the sample facial image, and reduces the feature loss.

[0074] Exemplarily, the candidate feature vector graph is input into the double-layer attention mechanism, and the candidate feature vector graph is divided into regions according to the height and width of the candidate feature vector graph to obtain a candidate feature vector subgraph. A candidate feature vector subgraph refers to a subgraph including at least part of the feature vector obtained after the candidate feature vector graph is divided. The double-layer attention mechanism can be used to focus on the key features in the candidate feature vector graph and ignore the edge features in the candidate feature vector graph to obtain a sample feature vector graph.

[0075] For example, the image partitioning region (such as S×S) is determined according to the height and width of the candidate feature vector image; and the candidate feature vector image is partitioned into non-overlapping regions according to the image partitioning region to obtain multiple candidate feature vector sub-graphs. H_f,W_f,C Input into the double-layer attention mechanism, the candidate feature vector graph is divided into 2×2 (i.e., graph partitioning area) non-overlapping areas, namely, candidate feature vector subgraphs. At this time, each candidate feature vector subgraph includes dimensional feature vector, where W_f represents the width of the candidate feature vector map; H_f represents the height of the candidate feature vector map; and C represents the number of output channels.

[0076] S220. Determine graph query data, graph key data and graph value data respectively according to the candidate feature vector subgraph, preset graph query weight, preset graph key weight and preset graph value weight, and determine the target feature vector subgraph from the candidate feature vector subgraph according to the graph query data and graph key data.

[0077] Among them, the embodiments of the present invention do not impose any restrictions on the sizes of preset graph query weights, preset graph key weights and preset graph value weights, which can be set by technicians based on experience, or repeatedly determined through a large number of experiments.

[0078] The graph query data may represent a region-level query matrix of the candidate feature vector graph, the graph key data may represent a region-level key matrix of the candidate feature vector graph, and the graph value data may represent a region-level value matrix of the candidate feature vector graph.

[0079] The target eigenvector subgraph refers to a candidate eigenvector subgraph containing a focused eigenvector. In an optional embodiment, the target eigenvector subgraph is determined from the candidate eigenvector subgraphs according to the graph query data and the graph key data, including: determining a regional association matrix of the candidate eigenvector subgraphs according to the graph query data and the graph key data; and selecting a preset number of target eigenvector subgraphs from the candidate eigenvector subgraphs according to the regional association matrix.

[0080] Among them, the region correlation matrix refers to a matrix composed of the correlation degrees between different candidate eigenvector subgraphs. The region correlation matrix includes the candidate region correlations between different candidate eigenvector subgraphs. The candidate region correlation refers to the correlation degree between any two candidate eigenvector subgraphs. The embodiment of the present invention does not impose any limitation on the number of preset graphs, which can be set by technicians based on experience or needs, or determined repeatedly through a large number of experiments.

[0081] Exemplarily, the regional correlation matrix can be determined by the following formula:

[0082] Q=X r W q ,K=X r W k ;

[0083] A r =Q(K) T ;

[0084] Among them, Q represents graph query data; K represents graph key data; X r Represents the feature vector graph obtained by reshaping the candidate feature vector graph, that is, W q Represents the preset graph query weight; W k Indicates the preset graph key weight; A r represents the region association matrix including the candidate region association; T represents transposition. r Each candidate eigenvector subgraph can be processed to obtain it.

[0085] Furthermore, according to I r =topklndex(A r ), select the top k (preset number of graphs) candidate regions with the largest associations from the region association matrix as the target region association, forming I r According to I r The target region association degree in is used, and the corresponding candidate feature vector subgraph is used as the target feature vector subgraph.

[0086] It can be understood that by introducing the regional correlation matrix and determining the target eigenvector subgraph according to the regional correlation matrix, the screening of candidate eigenvector subgraphs is realized, the participation of irrelevant eigenvectors in subsequent processing is reduced, resource waste is reduced, and the processing efficiency of the dual-layer attention mechanism is improved.

[0087] S230, determining a sample feature vector graph according to the target feature vector subgraph, the graph key data and the graph value data, and determining a sample prediction result of a sample training object in the sample facial image according to the sample feature vector graph.

[0088] The sample feature vector graph can be understood as a feature vector graph obtained by amplifying key feature values ​​in the candidate feature vector graph and reducing edge feature values ​​in the candidate feature vector graph.

[0089] In an optional embodiment, a sample feature vector graph is determined based on a target feature vector subgraph, graph key data, and graph value data, including: determining pixel key data based on the graph key data and the target feature vector subgraph, and determining pixel value data based on the graph value data and the target feature vector subgraph; determining a sample feature vector graph based on the graph query data, pixel key data, and pixel value data.

[0090] The pixel key data may represent a pixel-level key matrix of the target feature vector subgraph, and the pixel value data may represent a pixel-level value matrix of the target feature vector subgraph.

[0091] Exemplarily, the graph value data may be determined by the following formula:

[0092] V=X r W v ;

[0093] Wherein, V represents the image value data; W v Represents the preset graph value weight.

[0094] Exemplarily, the sample feature vector graph may be determined by the following formula:

[0095] K global =gather(K,I r ),V global =gather(V,I r );

[0096]

[0097] Among them, K global Represents pixel key data; gather means applying attention to the collected key-value pairs; V globalrepresents pixel value data; O represents sample feature vector graph; m is a custom constant; Softmax represents a normalized exponential function. If necessary, the sample feature vector graph includes -dimensional feature vector.

[0098] It can be understood that by introducing pixel key data and pixel value data, the sample feature vector graph is determined according to the graph query data, pixel key data and pixel value data, and it is possible to select the key feature vectors from the target feature vector subgraph from the pixel level dimension, increase the values ​​of the key feature vectors, and reduce the values ​​of the edge feature vectors, thereby improving the accuracy of the sample feature vector graph, and further improving the accuracy of the subsequent sample prediction results determined based on the sample feature vector graph.

[0099] The sample prediction result refers to the result obtained by predicting the sample training object in the sample facial image based on the facial emotion recognition model. Exemplarily, the sample prediction result includes sample candidate emotions and corresponding sample candidate probabilities. The sample candidate emotions represent the emotions that the sample training object may show as predicted by the facial emotion recognition model. The sample candidate probability refers to the predicted probability corresponding to the sample candidate emotions.

[0100] Exemplarily, the sample feature vector graph is input into a fully connected layer to obtain a sample prediction result, wherein the fully connected layer can be used to predict the facial image to obtain a result.

[0101] S240. Determine a model loss value of the facial emotion recognition model according to the sample prediction results of the sample training object and the corresponding sample actual labels, and train the facial emotion recognition model according to the model loss value.

[0102] Among them, the model loss value refers to the loss function value of the facial emotion recognition model.

[0103] Exemplarily, the sample candidate emotion corresponding to the largest sample candidate probability in the sample prediction result is used as the sample target emotion; and the model loss value is determined according to the sample target emotion and the sample actual label of the corresponding sample training object.

[0104] S250. Obtain current patrol data of the current patrol cycle from the master robot in the robot scheduling system.

[0105] Among them, the current patrol data is determined by the historical customer quantity obtained by the main robot in the previous patrol cycle and the preset customer quantity threshold;

[0106] S260, obtaining customer facial images of each customer to be served in the corresponding service area according to the current patrol data, and inputting the customer facial images into the trained facial emotion recognition model, and outputting candidate facial emotions of the customer to be served and corresponding candidate emotion probabilities;

[0107] S270: Determine the target facial emotion of the customer to be served from the candidate facial emotions according to the candidate emotion probability, and determine the target service customer and the corresponding target service strategy from the customers to be served according to the target facial emotion.

[0108] The embodiment of the present invention provides a robot scheduling scheme, which obtains a candidate feature vector graph by adding a sample facial image as input to a constructed facial emotion recognition model, and divides the candidate feature vector graph into regions according to the height and width of the candidate feature vector graph to obtain a candidate feature vector subgraph; wherein the facial emotion recognition model includes a facial feature generation network, a two-layer attention mechanism and a fully connected layer; according to the candidate feature vector subgraph, a preset graph query weight, a preset graph key weight and a preset graph value weight, graph query data, graph key data and graph value data are respectively determined, and according to the graph query data and the graph key data, a target feature vector subgraph is determined from the candidate feature vector subgraph; according to the target feature vector subgraph, the graph key data and the graph value data, a sample feature vector graph is determined, and according to the sample feature vector graph, a sample prediction result of a sample training object in the sample facial image is determined; according to the sample prediction result of the sample training object and the corresponding sample actual label, a model loss value of the facial emotion recognition model is determined, and according to the model loss value, the facial emotion recognition model is trained, thereby improving the training mechanism of the facial emotion recognition model. The above scheme processes the sample facial images input into the facial emotion recognition model to obtain a target feature map, thereby focusing on the key features in the sample facial images, thereby improving the accuracy of subsequent sample prediction results based on the target feature map; at the same time, by determining the model loss value based on the sample prediction results and the actual sample labels, the facial emotion recognition model is trained according to the model loss value, thereby improving the accuracy of model training.

[0109] Embodiment 3

[0110] This embodiment provides an optional example based on the above embodiment. It should be noted that for the parts not described in detail in the embodiment of the present invention, reference can be made to the descriptions of other embodiments.

[0111] In the existing technology, although intelligent robots play an important role in financial institutions, their effects and potential in practical applications still need to be further explored and improved. There are still problems such as robots are often in a waiting state during working hours in financial institutions and lack of active customer interaction, resulting in low customer attention, making it difficult to realize the value of operations and unable to effectively replace manual services.

[0112] Based on the above problems, the embodiment of the present invention proposes a robot scheduling solution. Specifically, based on the facial expression recognition method, the anxiously waiting customers (i.e., target service customers) are found, and the slave robots are asked about their needs first, providing the best solutions to improve customer satisfaction. The robots are designed to have an active welcoming function. When customers enter a financial institution, the main robot can take the initiative to greet and ask about their needs, and guide customers to the corresponding service area to handle business. The present invention effectively improves the utilization rate of robots in the application scenarios of existing financial institutions and reduces the workload of business personnel.

[0113] In the embodiment of the present invention, when a customer walks into a financial institution, the master robot at the door proactively asks the type of business to be handled and guides the customer to the corresponding service area. When the slave robot arrives at the patrol time point, it proactively identifies the emotions of the customer to be served, proactively asks the target service customer who shows negative emotions such as anger and anxiety about their needs, gives suggestions on business scenarios, and comforts them emotionally.

[0114] Exemplarily, the processing flow of the robot dispatching system includes: setting up a master robot and slave robots, determining current patrol data, performing emotion recognition on customers, dispatching slave robots to serve target customers, and collecting customer service evaluation data.

[0115] Exemplarily, setting up a master robot and a slave robot includes: setting up a master robot at the entrance of a financial institution, and setting up one (or more) slave robots in the service area (such as the customer waiting area and the self-service area). The master robot is mainly responsible for welcoming customers at the entrance and taking numbers and business guidance, that is, guiding customers to the corresponding service area. When a customer walks into the service area of ​​the financial institution, the slave robot in the service area can recognize the facial emotions of the customer to be served and provide corresponding services to the target service customer.

[0116] Exemplarily, determining the current patrol data includes: each service area of ​​the financial institution needs to be equipped with a slave robot. The slave robot needs to patrol regularly to check whether there are customers who need help. In the embodiment of the present invention, a dynamic patrol method is adopted, that is, the current patrol data is determined by the number of historical customers obtained by the master robot in the previous patrol cycle and a preset customer number threshold.

[0117] Exemplarily, the emotion recognition of customers includes: collecting customer images of customers to be served in the corresponding service area when the robot is patrolling, and determining the facial emotions of the customers to be served based on the customer images and the facial emotion recognition model. Specifically, firstly, the face in the customer image is separated using the target detection algorithm to obtain the customer facial image; the customer facial image is input into the trained facial emotion recognition model to obtain the candidate facial emotions and candidate emotion probabilities of the customer to be served, and then the target facial emotion of the customer to be served is determined according to the candidate emotion probabilities.

[0118] Exemplarily, scheduling a slave robot to serve a target service customer includes: taking a customer to be served with a negative facial emotion as a target service customer; asking the target service customer whether he needs help, calming the target service customer, providing a reasonable solution or contacting the lobby manager, that is, determining the target service strategy corresponding to the target service customer.

[0119] Exemplarily, collecting customer service evaluation data includes: the robot scheduling system records the service detailed data corresponding to the target service strategy (including conversation details, emotions, business numbers and service times); after receiving service from the robot, the target service customer is invited to comment on the service (such as 1 star for bad and 5 stars for very good); for 4-star or 5-star service detailed data, robot reinforcement learning is provided to facilitate subsequent management personnel to analyze and optimize the robot's responses.

[0120] Figure 3 1 is a system module diagram of a robot scheduling system provided by Embodiment 3 of the present invention. The robot scheduling system includes a robot configuration and scheduling module, an emotion recognition module and a system monitoring and maintenance module.

[0121] Exemplarily, the robot configuration and scheduling module includes a master-slave robot configuration submodule and a patrol strategy calculation submodule. Among them, the master-slave robot configuration submodule can be used to set and configure the roles, positions and patrol strategies of the master and slave robots. The master robot is responsible for greeting customers at the door of the financial institution, actively asking about the business that needs to be handled, and guiding customers to the corresponding service area; the slave robot is responsible for regular patrols and serving customers, and identifying customers' emotions, and for anxious or angry customers, voice is issued to appease. Among them, the patrol strategy calculation module can use the designed patrol time calculation method to dynamically adjust the patrol data of different patrol cycles, and trigger the patrol behavior of the slave robot according to the changes in the number of people in the financial institution.

[0122] Exemplarily, the emotion recognition module may include an analysis and recognition submodule and a decision submodule. The analysis and recognition submodule may be used in a video processing unit, a face recognition processing unit, and an emotion recognition unit. Among them, the video processing unit may extract the feature vector of the video frame in real time based on the collected video to perform emotion recognition analysis. The face recognition processing unit may adopt a target detection algorithm to construct a facial expression recognition method to recognize the face. The emotion recognition unit may classify the expression of the recognized face based on the designed facial expression method to recognize the customer's emotion. The decision submodule may determine the target service customer and determine the target service strategy corresponding to the target service customer. Exemplarily, the decision submodule may analyze and identify customers with negative emotions, identify customer information based on facial recognition, retrieve customer portraits through customer numbers, determine their service types, actively inquire about their needs, and make lovely sounds to appease the customer's emotions.

[0123] Exemplarily, the system monitoring and maintenance module may include a service record and comment submodule, an operation status monitoring submodule, and a system maintenance submodule. Among them, the service record and comment submodule can record the detailed data of each service provided by the robot, including the service number, service time, etc.; after the robot service, the target service customer is invited to comment on the service so that subsequent management personnel can analyze and optimize the robot's answers and service strategies. The operation status monitoring submodule can monitor the entire system in real time, including the robot's operation status, network connection status, and service request processing. The system maintenance submodule can maintain the robot scheduling system, such as robot firmware upgrades, system configuration adjustments, etc.

[0124] The embodiment of the present invention adopts computer vision technology to identify customer emotions on behalf of financial institution personnel and actively provide services, thereby solving the problem of low attention and poor results for customers by financial institutions, and improving the service quality of financial institutions to customers, optimizing human resource allocation, and enhancing brand image.

[0125] The robot scheduling solution provided by the embodiment of the present invention designs a patrol time calculation method and a facial expression recognition method, which improves the scheduling performance of the robot, strengthens the service initiative of the robot, reduces the burden of business services, and reduces the cost of employment; at the same time, the embodiment of the present invention enhances the customer's comfort and satisfaction by identifying the customer's emotions, thereby improving the customer experience; and the embodiment of the present invention adopts a dynamic patrol strategy to ensure that help can be provided in time when needed, thereby improving resource utilization and service efficiency; further, the technical solution provided by the embodiment of the present invention enhances the customer's trust and loyalty to the enterprise by providing high-quality customer service, and helps to form a good word-of-mouth effect.

[0126] It should be noted that the information collected in the embodiments of the present invention is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data comply with the relevant laws, regulations and standards of relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse; if the user chooses to refuse, the expert decision-making process will be entered.

[0127] Embodiment 4

[0128] Figure 4 This is a schematic diagram of the structure of a robot scheduling device provided in Embodiment 4 of the present invention. This embodiment can be applied to the situation of scheduling robots in financial institutions. The method can be executed by a robot scheduling device, which can be implemented in software and / or hardware and can be configured in an electronic device that carries the robot scheduling function.

[0129] like Figure 4 As shown, the device is configured in a slave robot in a robot scheduling system, and the device includes: a patrol data acquisition module 310, a model output module 320 and a service strategy determination module 330. Among them,

[0130] A patrol data acquisition module 310 is used to acquire current patrol data of a current patrol cycle from a master robot in the robot scheduling system; wherein the current patrol data is determined by the number of historical customers acquired by the master robot in a previous patrol cycle and a preset customer number threshold;

[0131] The model output module 320 is used to obtain the customer facial images of each customer to be served in the corresponding service area according to the current patrol data, and input the customer facial images into the trained facial emotion recognition model to output the candidate facial emotions of the customer to be served and the corresponding candidate emotion probabilities;

[0132] A service strategy determination module 330 is used to determine the target facial emotion of the customer to be served from the candidate facial emotions according to the candidate emotion probability, and determine the target service customer and the corresponding target service strategy from the customer to be served according to the target facial emotion;

[0133] The facial emotion recognition model is trained based on sample facial images of sample training objects and corresponding sample actual labels.

[0134] An embodiment of the present invention provides a robot scheduling solution, which obtains current patrol data of a current patrol cycle from a master robot in a robot scheduling system; wherein the current patrol data is determined by the number of historical customers obtained by the master robot in the previous patrol cycle and a preset customer number threshold; based on the current patrol data, customer facial images of each customer to be served in a corresponding service area are obtained, and the customer facial images are input into a trained facial emotion recognition model, and candidate facial emotions of the customer to be served and corresponding candidate emotion probabilities are output; based on the candidate emotion probabilities, target facial emotions of the customer to be served are determined from the candidate facial emotions, and based on the target facial emotions, target service customers and corresponding target service strategies are determined from the customers to be served; wherein the facial emotion recognition model is trained based on sample facial images of sample training objects and corresponding sample actual labels. In the above scheme, the robot scheduling system includes a master robot and a slave robot. The slave robot patrols and obtains customer facial images in the corresponding service area through the current patrol data obtained from the master robot. Subsequently, the target service customers that the slave robot needs to serve in the service area are determined based on the customer facial images and facial emotion recognition models, thereby improving the accuracy of robot scheduling; and, the current patrol data in the current patrol cycle in the embodiment of the present invention is predicted by the master robot based on the historical number of customers in the previous patrol cycle and a preset customer number threshold. The slave robot obtains the customer facial images of each customer to be served in the corresponding service area based on the current patrol data, thereby realizing dynamic adjustment of the current patrol data of the slave robot in the current patrol cycle, improving the effectiveness of obtaining customer facial images, thereby improving the accuracy of subsequent determination of target service customers based on customer facial images and facial emotion recognition models, and improving the timeliness of robot scheduling.

[0135] Optionally, the facial emotion recognition model is trained using the following devices:

[0136] A candidate subgraph determination module is used to input the sample facial image into the constructed facial emotion recognition model to obtain a candidate feature vector graph, and divide the candidate feature vector graph into regions according to the height and width of the candidate feature vector graph to obtain a candidate feature vector subgraph; wherein the facial emotion recognition model includes a facial feature generation network, a two-layer attention mechanism and a fully connected layer;

[0137] a target subgraph determination module, configured to determine graph query data, graph key data, and graph value data respectively according to the candidate feature vector subgraph, the preset graph query weight, the preset graph key weight, and the preset graph value weight, and to determine a target feature vector subgraph from the candidate feature vector subgraph according to the graph query data and the graph key data;

[0138] A prediction result determination module, configured to determine a sample feature vector graph according to the target feature vector subgraph, the graph key data and the graph value data, and determine a sample prediction result of a sample training object in the sample facial image according to the sample feature vector graph;

[0139] The model training module is used to determine the model loss value of the facial emotion recognition model according to the sample prediction results of the sample training object and the corresponding sample actual labels, and train the facial emotion recognition model according to the model loss value.

[0140] Optionally, the target subgraph determination module includes:

[0141] A region relevance matrix determining unit, used to determine the region relevance matrix of the candidate eigenvector subgraph according to the graph query data and the graph key data;

[0142] The target subgraph determining unit is used to select a preset number of target feature vector subgraphs from the candidate feature vector subgraphs according to the region association matrix.

[0143] Optionally, the prediction result determination module includes:

[0144] a pixel data determination unit, configured to determine pixel key data according to the graph key data and the target feature vector subgraph, and to determine pixel value data according to the graph value data and the target feature vector subgraph;

[0145] The sample feature vector graph determining unit is used to determine the sample feature vector graph according to the graph query data, the pixel key data and the pixel value data.

[0146] Optionally, the current patrol data is determined by the following means:

[0147] A quantity comparison module, used to compare the historical customer quantity with a preset customer quantity threshold;

[0148] The current data determination module is used to determine the current patrol data of the slave robot in the current patrol cycle according to the comparison result.

[0149] Optionally, if the current patrol data includes the current patrol times and the current patrol time, the current data determination module includes:

[0150] A basic number determination unit, configured to obtain a basic patrol number of the slave robot in a current patrol cycle if the comparison result is that the number of historical customers is greater than or equal to a preset customer number threshold;

[0151] A patrol times determination unit is used to determine the current patrol times of the slave robot in the current patrol cycle according to the basic patrol times, the historical customer numbers, the preset customer number threshold and the preset frequency increase coefficient, and to determine the current patrol time according to the current patrol times and the duration of the current patrol cycle; wherein the current patrol times is an integer and is greater than or equal to the basic patrol times.

[0152] Optionally, the service policy determination module 330 includes:

[0153] A target customer determination unit, configured to determine the customer to be served whose target facial emotion is a negative facial emotion as a target service customer;

[0154] The target strategy determination unit is used to determine the business type of the target service customer, and determine the target service strategy of the target service customer according to the business type and the negative facial emotion.

[0155] The robot scheduling device provided in the embodiment of the present invention can execute the robot scheduling method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing each robot scheduling method.

[0156] In the technical solution of the present invention, the collection, storage, use, processing, transmission, provision and disclosure of the historical customer number, sample facial images, sample actual labels and customer facial images involved all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0157] Embodiment 5

[0158] Figure 5 4 is a schematic diagram of the structure of an electronic device for implementing a robot scheduling method provided by Embodiment 5 of the present invention. The electronic device 410 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0159] like Figure 5As shown, the electronic device 410 includes at least one processor 411, and a memory connected to the at least one processor 411 in communication, such as a read-only memory (ROM) 412, a random access memory (RAM) 413, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 411 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 412 or the computer program loaded from the storage unit 418 to the random access memory (RAM) 413. In the RAM 413, various programs and data required for the operation of the electronic device 410 can also be stored. The processor 411, the ROM 412, and the RAM 413 are connected to each other via a bus 414. An input / output (I / O) interface 415 is also connected to the bus 414.

[0160] Multiple components in the electronic device 410 are connected to the I / O interface 415, including: an input unit 416, such as a keyboard, a mouse, etc.; an output unit 417, such as various types of displays, speakers, etc.; a storage unit 418, such as a disk, an optical disk, etc.; and a communication unit 419, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 419 allows the electronic device 410 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0161] The processor 411 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 411 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 411 executes the various methods and processes described above, such as a robot scheduling method.

[0162] In some embodiments, the robot scheduling method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 418. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 410 via the ROM 412 and / or the communication unit 419. When the computer program is loaded into the RAM 413 and executed by the processor 411, one or more steps of the robot scheduling method described above may be performed. Alternatively, in other embodiments, the processor 411 may be configured to execute the robot scheduling method in any other appropriate manner (e.g., by means of firmware).

[0163] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0164] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0165] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0166] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0167] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0168] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0169] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0170] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A robot scheduling method, characterized in that: Applied to a slave robot in a robot scheduling system, the method comprises: Acquire current patrol data of a current patrol cycle from a master robot in the robot dispatching system; wherein the current patrol data is determined by a historical number of customers acquired by the master robot in a previous patrol cycle and a preset customer number threshold; According to the current patrol data, obtain the customer facial images of each customer to be served in the corresponding service area, input the customer facial images into the trained facial emotion recognition model, and output the candidate facial emotions of the customer to be served and the corresponding candidate emotion probabilities; Determine the target facial emotion of the customer to be served from the candidate facial emotions according to the candidate emotion probabilities, and determine the target service customer and the corresponding target service strategy from the customers to be served according to the target facial emotion; The facial emotion recognition model is trained based on sample facial images of sample training objects and corresponding sample actual labels.

2. The method according to claim 1, characterized in that The facial emotion recognition model is trained in the following way: Inputting the sample facial image into the constructed facial emotion recognition model to obtain a candidate feature vector map, and dividing the candidate feature vector map into regions according to the height and width of the candidate feature vector map to obtain a candidate feature vector sub-map; wherein the facial emotion recognition model includes a facial feature generation network, a two-layer attention mechanism and a fully connected layer; Determine graph query data, graph key data, and graph value data respectively according to the candidate feature vector subgraph, the preset graph query weight, the preset graph key weight, and the preset graph value weight, and determine a target feature vector subgraph from the candidate feature vector subgraph according to the graph query data and the graph key data; Determine a sample feature vector graph according to the target feature vector subgraph, the graph key data and the graph value data, and determine a sample prediction result of a sample training object in the sample facial image according to the sample feature vector graph; According to the sample prediction results of the sample training objects and the corresponding sample actual labels, the model loss value of the facial emotion recognition model is determined, and according to the model loss value, the facial emotion recognition model is trained.

3. The method according to claim 2, characterized in that The step of determining a target feature vector subgraph from the candidate feature vector subgraphs according to the graph query data and the graph key data comprises: Determining a regional association matrix of the candidate eigenvector subgraph according to the graph query data and the graph key data; According to the region association matrix, a preset number of target eigenvector subgraphs are selected from the candidate eigenvector subgraphs.

4. The method according to claim 2, characterized in that: The step of determining a sample feature vector graph according to the target feature vector subgraph, the graph key data and the graph value data comprises: Determining pixel key data based on the graph key data and the target eigenvector subgraph, and determining pixel value data based on the graph value data and the target eigenvector subgraph; A sample feature vector graph is determined based on the graph query data, the pixel key data and the pixel value data.

5. The method according to claim 1, characterized in that The current patrol data is determined in the following manner: Comparing the historical number of customers with a preset customer number threshold; According to the comparison result, the current patrol data of the slave robot in the current patrol cycle is determined.

6. The method according to claim 5, characterized in that If the current patrol data includes the current patrol times and the current patrol time, then the current patrol data of the slave robot in the current patrol cycle is determined according to the comparison result, including: If the comparison result is that the number of historical customers is greater than or equal to the preset customer number threshold, then obtaining the basic patrol times of the slave robot in the current patrol cycle; The current patrol number of the slave robot in the current patrol cycle is determined based on the basic patrol number, the historical customer number, the preset customer number threshold and the preset frequency increase coefficient, and the current patrol time is determined based on the current patrol number and the duration of the current patrol cycle; wherein the current patrol number is an integer and the current patrol number is greater than or equal to the basic patrol number.

7. The method according to any one of claims 1 to 6, characterized in that The step of determining a target service customer and a corresponding target service strategy from the customers to be served according to the target facial emotion includes: The customer to be served whose target facial emotion is a negative facial emotion is taken as a target service customer; The business type of the target service customer is determined, and a target service strategy for the target service customer is determined based on the business type and the negative facial emotion.

8. A robot scheduling device, characterized in that: A slave robot configured in a robot scheduling system, the device comprising: A patrol data acquisition module, used to acquire current patrol data of a current patrol cycle from a master robot in the robot scheduling system; wherein the current patrol data is determined by the number of historical customers acquired by the master robot in a previous patrol cycle and a preset customer number threshold; A model output module is used to obtain customer facial images of each customer to be served in the corresponding service area according to the current patrol data, and input the customer facial images into the trained facial emotion recognition model to output candidate facial emotions of the customer to be served and corresponding candidate emotion probabilities; A service strategy determination module, used to determine the target facial emotion of the customer to be served from the candidate facial emotions according to the candidate emotion probability, and determine the target service customer and the corresponding target service strategy from the customers to be served according to the target facial emotion; The facial emotion recognition model is trained based on sample facial images of sample training objects and corresponding sample actual labels.

9. An electronic device, characterized in that: include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement a robot scheduling method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a robot scheduling method as described in any one of claims 1 to 7 is implemented.

Images