Service consumption intelligent robot for preventing service failure and working method

Abstract

The invention relates to the technical field of intelligent robots, in particular to a service consumption intelligent robot for preventing service failure and a working method, and aims at collecting data in multiple dimensions by using a 3D visual sensor, a pressure sensor and an interactive interface log and analyzing and predicting a potential service failure risk. If an operation confusion risk or a bearing insufficiency risk is detected, the mechanical arm is triggered to act in advance, the mechanical arm guides swing in a high-frequency and small-amplitude mode during operation confusion, the horizontal platform is unfolded and hydraulic buffering is started during bearing insufficiency, active intervention is conducted before service failure occurs, and the service failure probability is greatly reduced; the limitation that the prior art depends on post remedy is broken through, and a complete'monitoring-prediction-intervention-optimization 'prevention closed loop is constructed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent robots, and particularly relates to an intelligent service consumption robot for preventing service failures and a working method thereof. Background Art

[0002] With the rapid development of artificial intelligence technology, intelligent robots are increasingly widely used in many fields. Especially in the service industry, intelligent robots have gradually become important service carriers. However, in the actual application process, the service quality of intelligent robots faces many challenges.

[0003] In the service scenario of intelligent robots, service failures often occur due to technical limitations and complex and changeable service environments. When service failures occur, most of the existing technologies rely on "post-remedial" measures. For example, when problems occur, such as when a person cannot operate the robot interface and a reminder of operation failure appears, and then control is achieved through the movement of the robot's manipulator. This can only perform the function of post-remedial.

[0004] In the control process of the robot's manipulator, such as Chinese Patent Application No. 202411953555.8, the publication date is March 18, 2025, and the classification number is B25J9 / 16. The robot includes a main control servo, m slave control manipulator arms, and a photographing device. The photographing device is arranged on at least one of the slave control manipulator arms. The method includes: obtaining environmental image data photographed by the photographing device; inputting the environmental image data into a target model for processing to obtain control instructions for each of the m slave control manipulator arms. The target model includes a first network and a second network. The first network is used to calculate joint motion data corresponding to each joint in the target manipulator arm. The target manipulator arm is at least one of the m slave control manipulator arms. The second network is used to perform mapping processing based on the joint motion data corresponding to each joint in the target manipulator arm to obtain control instructions for each of the m slave control manipulator arms. The above can improve the accuracy and efficiency of robot control.

[0005] The robots in the above literature mainly make adjustments after problems occur during service. For consumer service robots, during the service process, the user's attention is easily concentrated on the operation interface. Once there are problems with the operation or the waiting time is too long, anxiety will be generated. Existing intelligent robots lack effective ways to divert the user's attention and cannot actively improve the user experience during the service process. Taking the hotel check-in scenario as an example, if the intelligent robot is slow in handling the check-in procedures, the user can only keep staring at the operation interface and waiting, and there are no measures to relieve anxiety during this period. This will greatly reduce the user's satisfaction with the service. In fact, the manipulator can be controlled only when problems occur currently, thus reducing the user experience. Summary of the Invention

[0006] The present invention provides a service - consuming intelligent robot for preventing service failures and a working method thereof, which can control the robot's manipulator in advance according to the current reading situation of the robot, so as to prevent failures in advance and provide a service experience for users.

[0007] On the one hand, the present invention provides a working method for a service - consuming intelligent robot for preventing service failures, including the following steps: Step S1: Real - time monitor the user's behavior characteristics through a scanning area; the user's behavior characteristics include four categories: capturing the user's limb movements, item - carrying status, facial expression characteristics, and the load - bearing status of the robotic arm; determine the user's typical behavior pattern through a clustering algorithm, and update the behavior - feature weights of the four categories in real - time through a learning method to determine the user's behavior pattern; Step S2: Analyze the user's needs according to the user's behavior pattern and predict the potential service - failure risk type; Step S3: Trigger the corresponding robotic - arm action pattern according to the predicted potential service - failure risk type.

[0008] In the above method, by scanning the user's behavior characteristics, such as the user's limb movements, item - carrying status, facial expression characteristics, and the load - bearing status of the robotic arm, etc., and determining the final behavior of each category through clustering analysis for the four categories, and updating the behavior - feature weights of the four categories in real - time through a learning method, so as to determine the user's behavior pattern. For example, when the current user's limb movement is scratching the head and the facial expression is distressed, and the item - carrying status is good, the weights of the categories of the user's limb movements and facial expressions are increased a little, so as to determine the user's behavior pattern as not knowing how to operate. Thus, the user's needs for guidance can be analyzed through the user's behavior pattern, and the action pattern of the robotic arm is determined as a guiding operation by predicting the potential service - failure risk type of this kind. In this way, the user's failure situation can be predicted in advance, and intervention can be carried out in advance, so as to ensure a good user experience, form a complete prevention closed - loop of "monitoring - prediction - intervention - optimization", realize the transformation of the service process from "post - event remedy" to "pre - event prevention", and establish a new type of service relationship of human - machine collaboration.

[0009] Further, in step S1, the monitoring of the user's behavior specifically includes the following steps: c1: Continuously capture the user's limb movements, item - carrying status, and facial expression characteristics by using a 3D vision sensor, and detect the mechanical - arm load - weight distribution data in real - time through a pressure sensor, and record the user's operation trajectory in detail with the help of the interactive - interface log; Use the 3D vision sensor to record the user's hand click and slide actions, the size, shape, and walking posture of the carried items when operating the robot's interactive interface. The pressure sensor records the weight values borne by each part of the robotic arm at different time periods. The operation trajectory is the order of clicked buttons and page - switching records; c2. Clean the collected raw data, remove the error data and outliers caused by sensor failures or other interference factors, and then standardize different types of data; The 3D vision sensor captures limb movement data that does not conform to the normal human movement logic, which can be regarded as outliers and excluded. The standardization process is to unify the range of the weight data of the pressure sensor and the limb movement amplitude data detected by the 3D vision sensor.

[0010] c3. Use the clustering algorithm to analyze the preprocessed data, and continuously adjust the parameters of the clustering algorithm to determine the optimal clustering result through evaluation indicators; c4. Determine the weights of the four clustering results and determine the user behavior pattern through dynamic learning.

[0011] The above method classifies user behavior data with a series of similar operation trajectories, similar limb movements and carried item states into one category, and determines the optimal clustering result through cluster analysis to ensure the rationality and accuracy of the cluster division; and adjusts the weights of the clustering results through dynamic learning after obtaining the clustering results, thereby determining the user behavior pattern.

[0012] Further, the step c4 further includes dynamic learning optimization, and the dynamic learning optimization is to establish a user behavior baseline model, distinguish typical behavior patterns through cluster analysis; use the FTRL online learning algorithm to update the behavior feature weights in real time; trigger an active calibration mechanism for abnormal data.

[0013] The above settings update the behavior feature weights by means of the FTRL prior learning algorithm, so as to ensure that the finally obtained user behavior pattern can more accurately determine the user behavior.

[0014] Further, in the step S2 of predicting the potential service failure risk, when it is detected that the interface operation sequence deviates from the standard process, it is determined as an operation confusion risk; when it is detected that the user's eye wandering rate exceeds the threshold, it is determined as an operation confusion risk; when the total volume of the scanned items is greater than 80% of the maximum load capacity of the robotic arm, it is determined as a load insufficiency risk; when the variance of the item weight distribution is greater than the safety threshold, it is determined as a load insufficiency risk; when it is detected that the user has a carried item and the user's limb is biased towards one side of the item and the expression is in a painful state, it is determined as a carrying inconvenience risk.

[0015] The above settings can analyze and confirm the risks through the detected different situations, which is convenient for subsequent corresponding measures to be taken for the risks.

[0016] Further, the step S3 of triggering the corresponding robotic arm action mode according to the risk type includes: b1. For the risk of operational confusion, the robot arm is controlled to perform high-frequency and small-amplitude guiding swings with a swing frequency of 5-8 Hz; b2. For the risk of insufficient load, control the robot arm to deploy the horizontal platform and start the hydraulic buffer; b3. To avoid the risk of inconvenience in carrying, control the robotic arm to move closer to the location of the item.

[0017] The above settings can better help users by controlling the robotic arm to perform corresponding operations for different risks, thereby preventing users from having a bad user experience due to inadequate service.

[0018] Furthermore, step S3 also includes: shifting the user's attention focus by dynamic posture changes by applying a visual saliency model to predict the user's line of sight, and when it is detected that the user continues to look at the wrong area, controlling the robotic arm to perform sudden actions and guiding the line of sight to shift with the help of lights.

[0019] The above arrangement enables the robot arm to provide guidance when the user continues to focus on the wrong area.

[0020] Furthermore, the clustering algorithm includes: Data points are divided into Cluster In the process, the sum of the squares of the distances from each data point to the center of the cluster to which it belongs is minimized. Let the data point Belongs to cluster , cluster The center is , then the objective function is: ; Constantly adjust the K value of the clustering algorithm and determine the optimal clustering result through the evaluation index, the evaluation index is the silhouette coefficient , for a data point , the silhouette coefficient is defined as ; in yes The average distance to other data points in the same category, yes The average distance to the nearest data point in other categories; the silhouette coefficient of the entire data set is the average of the silhouette coefficients of all data points ; Select The largest K value is used as the optimal clustering number to determine the objective function and obtain the clustering result.

[0021] The above settings continuously adjust the K value of the clustering algorithm so that the average value of the silhouette coefficient of the data points is the largest, thereby ensuring that the silhouette effect is good and more accurate.

[0022] On the other hand, the present invention provides a service consumption intelligent robot for preventing service failures, including: a robot main body and a base provided at the bottom thereof, an interaction interface is provided on the outer part of the robot main body, robotic arms are symmetrically arranged on both sides of the robot main body, and the robotic arms are used to perform preset posture actions according to the service state to divert the user's attention. A scanning area is provided on the outer part of the robot main body, and the scanning area is used to identify user characteristics. A control module is provided inside the robot main body, which is used to control the action mode switching of the robotic arms according to the input signal of the scanning area.

[0023] With the above settings, by scanning user behavior characteristics, such as user limb movements, item carrying status, facial expression characteristics, and the load-bearing status of the robotic arms, etc., and determining the final behavior of each category through cluster analysis for four categories, and updating the weight of the behavior characteristics of the four categories in real time through a learning method, so as to determine the user behavior pattern. For example, when the current user limb movement is scratching the head and the facial expression is distressed, and the item carrying status is good, the weight of the category of user limb movement and facial expression is increased a little, so as to determine the user behavior pattern as not knowing how to operate. Thus, the user's needs can be analyzed through the user behavior pattern to obtain guidance, and the action mode of the robotic arm is determined as a guiding operation by predicting the potential service time limit risk type of this category. In this way, the user failure situation can be predicted in advance, and intervention can be carried out in advance, so as to ensure a good user experience effect, form a complete prevention closed-loop of "monitoring - prediction - intervention - optimization", realize the transformation of the service process from "post-remedy" to "prevention", and establish a new type of service relationship of human-machine collaboration.

[0024] Further, when the robotic arm detects that the printing task is completed, it automatically extends and presents a bill delivery posture. When the scanning area identifies that the user is carrying multiple items, the robotic arm automatically adjusts to a horizontal load-bearing posture.

[0025] With the above settings, the robotic arm can be conveniently controlled through various states of the robotic arm. Description of the Drawings

[0026] Figure 1 is the system flow chart of the present invention.

[0027] Figure 2 is the structural schematic diagram of the present invention.

[0028] Figure 3 is the display interface diagram of the present invention. Detailed Embodiments

[0029] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are given for purposes of illustration and description, and are not exhaustive or limit the present invention to the disclosed form. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles of the present invention and its practical applications, and to enable those of ordinary skill in the art to understand the present invention and thus design various embodiments with various modifications suitable for specific purposes.

[0030] Embodiment: Please refer to Figure 1 - Figure 3 , the present invention provides a technical solution: a service consumption intelligent robot for preventing service failures, including: a robot main body 2 and a base 1 provided at its bottom, an interaction interface is provided outside the robot main body 2, robotic arms are symmetrically provided on both sides of the robot main body 2, and the robotic arms are used to execute preset posture actions according to the service state to divert the user's attention. A scanning area 3 is provided outside the robot main body 2, and the scanning area 3 is used to identify the user's needs. A control module is provided inside the robot main body 2 for controlling the switching of the action mode of the robotic arm according to the input signal of the scanning area.

[0031] When the robotic arm detects that the printing task is completed, it automatically extends and presents a bill delivery posture. When the scanning area identifies that the user is carrying multiple items, the robotic arm automatically adjusts to a horizontal carrying posture. When the robotic arm detects that the user cannot operate, it will adjust to a guiding state, and when the user is carrying heavy items, the control module controls the robotic arm to approach the items. In this embodiment, the control module controls the robotic arm to move to the target position or perform corresponding operations, which is the existing robotic arm control principle and is not described in detail here as it is prior art.

[0032] As Figure 1 shown, a working method of a service consumption intelligent robot for preventing service failures includes the following steps: Step S1, real-time monitor the user's behavior characteristics through the scanning area; the user's behavior characteristics include four categories: capturing the user's limb movements, item carrying status, facial expression characteristics, and the load-bearing status of the robotic arm; determine the user's typical behavior pattern through a clustering algorithm, and update the behavior characteristic weights of the four categories in real time through a learning method to determine the user's behavior pattern; Step S2, analyze the user's needs according to the user's behavior pattern and predict the potential service failure risk type; Step S3, trigger the corresponding robotic arm action mode according to the predicted potential service failure risk type.

[0033] In step S1, the monitoring of the user's behavior specifically includes the following steps: Step a1, capturing the user's body movements, item carrying status and facial expressions through a 3D vision sensor; in this embodiment, the 3D vision sensor capturing the user's body movements, item carrying status and facial expressions is an existing visual recognition principle, specifically, the specific user's body movements, item carrying status and facial expressions are determined by comparing the visually scanned image with a preset standard image.

[0034] Step a2, detecting the weight distribution of the robot arm through a pressure sensor; Step a3: Record user operation traces through the interactive interface log.

[0035] In one embodiment, the user behavior baseline model establishment step in step S1 is: c1. Use 3D vision sensors to continuously capture the user's body movements, item carrying status, and facial features, and use pressure sensors to detect the weight distribution data of the robotic arm in real time, and use the interactive interface log to record the user's operation trajectory in detail; The 3D vision sensor is used to record the user's hand clicks and sliding movements when operating the robot's interactive interface, the size and shape of the objects carried, and the walking posture. The pressure sensor records the weight values borne by each part of the robot arm in different time periods. The operation track is the sequence of buttons clicked and the page switching record. c2. Clean the collected raw data, remove erroneous data and outliers caused by sensor failure or other interference factors, and then standardize different types of data; If the 3D vision sensor captures body movement data that does not conform to the normal human movement logic, it can be regarded as an outlier and removed. The normalization process is to unify the weight data of the pressure sensor and the body movement amplitude data detected by the 3D vision sensor. For example, if the vision sensor captures the movement of other people or objects behind the user, it is judged as body movement data that does not conform to the normal human movement logic and is regarded as an outlier.

[0036] c3. Use clustering algorithms to analyze the preprocessed data, and continuously adjust the parameters of the clustering algorithm to determine the optimal clustering results through evaluation indicators; Classify a series of user behavior data with similar operation trajectories, body movements, and similar items carried into one category, and determine the optimal clustering result by evaluating the silhouette coefficient to ensure the rationality and accuracy of cluster division; Taking the K-Means clustering algorithm as an example, its goal is to Data points are divided into Cluster In the process, the sum of the squares of the distances from each data point to the center of the cluster to which it belongs is minimized. Belongs to cluster , clusters The center of is ; Continuously adjust the K value of the clustering algorithm, and determine the optimal clustering result through evaluation metrics. The commonly used evaluation metric is the silhouette coefficient For a data point , its silhouette coefficient is defined as ; Among them is the average distance from to other data points in the same category, is the average distance from to data points in the nearest other category. The silhouette coefficient of the overall data set is the average of the silhouette coefficients of all data points. By trying different K values, calculate the corresponding silhouette coefficients, and select the K value that makes

[0037] c4. Conduct in-depth analysis on each clustering cluster, extract the typical behavior model that can represent the core characteristics of the cluster, and conduct detailed description and definition on the determined typical behavior pattern to form the user behavior baseline model.

[0038] In a certain cluster, it is found that most users follow a fixed order of first clicking on a specific function button and then placing an item when operating the robot, and the limb movements are stable reaching and placing movements. This can be determined as a typical behavior pattern; conduct detailed description and definition on the determined typical behavior pattern to form the basic framework of the user behavior baseline model, which is used to distinguish normal behavior from abnormal behavior in the future, and provide a benchmark for dynamic learning and optimization.

[0039] For the limb movement cluster, the mean vector of all limb movement vectors in the cluster can be calculated as the typical limb movement pattern. Suppose the cluster has limb movement vectors , then the typical limb movement pattern .

[0040] For the pressure sensor data cluster, the mean vector and standard deviation vector of the load weight distribution vector in the cluster can be calculated. The mean vector represents the average load weight distribution of the cluster, and the standard deviation vector reflects the degree of dispersion of the data.

[0041] For the cluster of interactive interface operation trajectories, the operation event sequence with the highest occurrence frequency can be found as the typical operation trajectory pattern. If in the cluster the operation event sequence has the highest number of occurrences, then it can be used as the typical operation trajectory pattern of this cluster.

[0042] For a certain cluster, it can be described as: when the user presents limb movements characterized by and the weight distribution of the load carried by the robotic arm is close to , and the operation trajectory follows pattern, it belongs to this typical user behavior.

[0043] c4. Determine the weights of the clustering results through dynamic learning and determine the user behavior pattern. In this embodiment, the FTRL online learning algorithm is used to update the weights of behavior features in real time; an active calibration mechanism is triggered for abnormal data.

[0044] For example, for the clustering results, initially preset their respective feature weights K1, K2, K3. When the clustering results are detected, and the feature weight value corresponding to a clustering result with a large difference from the preset evaluation criteria is increased. The increase value is preset, so that a clustering result with a large difference from the preset evaluation criteria can be better reflected, and the user behavior pattern can be determined with this clustering result as the main influencing factor. For example, for a user, for the operation clustering, it is currently an operation failure, but for the limb movements, such as carrying multiple items or the items being too heavy. If the clustering result of the items exceeds 60% of the preset clustering result, and the clustering result of the operation error interface exceeds 10% of the preset clustering result, and the clustering result of the limb movement cluster exceeds 0% of the preset clustering result, then the feature weight of the item clustering result is increased by 10, so that the entire user behavior is determined as a risk of insufficient load bearing.

[0045] In the prediction of potential service failure risks in step S2, when it is detected that the interface operation sequence deviates from the standard process, it is determined as an operation confusion risk; when it is detected that the user's eye gaze deviation rate exceeds the threshold, it is determined as an operation confusion risk; when the total volume of the scanned items is greater than 80% of the maximum load capacity of the robotic arm, it is determined as a risk of insufficient load bearing; when the variance of the item weight distribution is greater than the safety threshold, it is determined as a risk of insufficient load bearing; when it is detected that the user is carrying an item and the user's limb is biased towards one side of the item and the expression is in a painful state, it is determined as a risk of inconvenient carrying.

[0046] In step S3, triggering the corresponding robotic arm action mode according to the risk type includes: b1. For the risk of operation confusion, control the robotic arm to perform high-frequency small-amplitude guiding swings with a swing frequency of 5 - 8 Hz.

[0047] b2. For the risk of insufficient load-bearing, control the robotic arm to execute the horizontal platform deployment and activate the hydraulic buffer.

[0048] b3. For the risk of inconvenient carrying, control the robotic arm to approach the location where the item is located.

[0049] The working principle of the present invention: By scanning user behavior characteristics, such as user body movements, item carrying states, facial expression characteristics, and the load-bearing state of the robotic arm, etc., and determining the final behavior of each category through cluster analysis for four categories, and updating the weights of the behavior characteristics of the four categories in real time through a learning method, so as to determine the user behavior pattern. For example, when the current user body movement is scratching the head and the facial expression is distressed, and the item carrying state is good, the weight of the category of user body movement and facial expression is adjusted higher, so as to determine the user behavior pattern as not knowing how to operate. Thus, the user's needs for guidance can be analyzed through the user behavior pattern, and the action pattern of the robotic arm can be determined as a guiding operation by predicting the potential service timeliness risk type of this category. In this way, the user's failure situation can be predicted in advance, and intervention can be carried out in advance, so as to ensure a good user experience effect, form a complete prevention closed-loop of "monitoring - prediction - intervention - optimization", realize the transformation of the service process from "post-remedy" to "prevention", and establish a new type of service relationship of human-machine collaboration.

Claims

1. A working method of a service consumption intelligent robot for preventing service failure, characterized in that, Including: Step S1: Monitor the user's behavioral characteristics in real time through the scanning area; The user's behavioral characteristics include four categories: capturing the user's limb movements, item-carrying status, facial expression characteristics, and the load-bearing status of the robotic arm; determining the user's typical behavioral patterns through a clustering algorithm, and determining the user's behavioral patterns by updating the weight of the behavioral characteristics of the four categories in real time through a learning method; Step S2: Analyze the user's needs based on the user's behavioral patterns and predict the types of potential service failure risks; Step S3: Trigger the corresponding robotic arm action mode according to the predicted types of potential service failure risks.

2. The working method of a service consumption intelligent machine for preventing service failure according to claim 1, characterized in that: In step S1, the monitoring of the user's behavior specifically includes the following steps: c1: Continuously capture the user's limb movements, item-carrying status, and facial expression characteristics using a 3D vision sensor, and detect the weight distribution data of the robotic arm in real time through a pressure sensor, and record the user's operation trajectory in detail with the help of the interactive interface log; Use the 3D vision sensor to record the clicking and sliding actions of the user's hand when operating the robot interactive interface, the size and shape of the carried items, and the walking posture. The pressure sensor records the weight values borne by each part of the robotic arm at different time periods. The operation trajectory is the order of clicked buttons and page switching records; c2: Clean the collected raw data, remove the incorrect data and outliers caused by sensor failures or other interference factors, and then standardize the data of different types; The limb movement data captured by the 3D vision sensor that does not conform to the normal human movement logic can be regarded as outliers and removed. The standardization process is to unify the range of the weight data of the pressure sensor and the limb movement amplitude data detected by the 3D vision sensor. c3: Analyze the preprocessed data using a clustering algorithm, and continuously adjust the parameters of the clustering algorithm to determine the optimal clustering result through evaluation indicators; c4: Determine the weights of the four clustering results through dynamic learning and determine the user's behavioral patterns.

3. The working method of an intelligent service consumption robot for preventing service failures according to claim 2, characterized in that, The step c4 also includes dynamic learning optimization, which is to establish a user behavior baseline model and distinguish typical behavioral patterns through cluster analysis; use the FTRL online learning algorithm to update the behavioral feature weights in real time; trigger an active calibration mechanism for abnormal data.

4. The working method of a service consumption intelligent robot for preventing service failure according to claim 1, characterized in that, In the prediction of potential service failure risks in step S2, when it is detected that the interface operation sequence deviates from the standard process, it is determined as an operation confusion risk; when it is detected that the user's eye wandering rate exceeds the threshold, it is determined as an operation confusion risk; when the total volume of scanned items is greater than 80% of the maximum load capacity of the robotic arm, it is determined as an underload risk; when the variance of the item weight distribution is greater than the safety threshold, it is determined as an underload risk; when it is detected that the user is carrying an item and the user's limb is biased towards one side of the item and the facial expression is in a painful state, it is determined as an inconvenient carrying risk.

5. The working method of a service consumption intelligent robot for preventing service failures according to claim 1, characterized in that: In step S3, triggering the corresponding robotic arm action mode according to the risk type includes: b1: For the operation confusion risk, control the robotic arm to execute a high-frequency small-amplitude guiding swing, and the swing frequency is 5 - 8 Hz; b2: For the underload risk, control the robotic arm to execute a horizontal platform expansion and start the hydraulic buffer; b3. For the risk of inconvenience in carrying, control the robotic arm to approach the location where the item is located.

6. The work of a service consumption intelligent robot for preventing service failure according to claim 5 A method of making, characterized in that, The step S3 further includes: predicting the landing point of the user's line of sight by using a visual saliency model in a way of transferring the user's attention focus through dynamic posture changes, and when it is detected that the user continuously gazes at the wrong area, controlling the robotic arm to perform a sudden movement and cooperating with the light to guide the line of sight to transfer.

7. The working method of an intelligent service consumption robot for preventing service failures according to claim 1, characterized in that, The clustering algorithm includes: dividing data points into clusters such that the sum of the squared distances from each data point to the center of its belonging cluster is minimized. Let the data point belong to the cluster , and the center of the cluster be . Then the objective function is: ; Continuously adjust the K value of the clustering algorithm and determine the optimal clustering result through evaluation metrics. The evaluation metric is the silhouette coefficient , for a data point , the silhouette coefficient is defined as ; Among them is the average distance to other data points in the same category, is the average distance to the data points in the nearest other category; the silhouette coefficient of the overall data set is the average of the silhouette coefficients of all data points ; Select the K value that makes the largest as the optimal number of clusters to determine the objective function and obtain the clustering result.

8. The intelligent service consumption robot for preventing service failure according to claim 1, comprising: The robot body and a base provided at the bottom thereof, an interaction interface is provided on the outside of the robot body, robotic arms are symmetrically provided on both sides of the robot body, the robotic arms are used to perform preset posture actions according to the service state to transfer the user's attention, a scanning area is provided on the outside of the robot body, the scanning area is used to identify the user's needs, and a control module is provided inside the robot body for controlling the switching of the action mode of the robotic arm according to the input signal of the scanning area.

Images