A multi-task scheduling method and device for an artificial intelligence robot

Abstract

The present application proposes a multi-task scheduling method and device for an artificial intelligence robot. The first user's trip is divided into individual trips and group trips according to the trip scale. When the first user's matching degree with historical users meets the preset requirements, the first user's activity information is obtained according to the first user's historical portrait. The first user's image and the first group's image are combined to obtain the first user's portrait for the group's travel mode. The first user's portrait can better portray the group's travel preferences. When tasks are assigned, the robot's real-time location and the first user's feedback information are combined to update the tasks in real time, so that the robot's services can better meet the user's needs. The present application proposes a multi-task scheduling method and device for an artificial intelligence robot. The robot's services can better meet the user's actual needs.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent robots, and in particular relates to a multi-task scheduling method and device for an artificial intelligence robot. Background Art

[0002] With the development of artificial intelligence technology, intelligent robots have been applied in various scenarios, but existing intelligent robots can only mechanically provide corresponding services based on users' search requests, and cannot provide targeted services based on users' historical data or intelligent analysis data. Summary of the invention

[0003] The purpose of the present invention is to provide a multi-task scheduling method and device for an artificial intelligence robot to solve the technical problem in the prior art that it is impossible to provide targeted services to users based on their historical data or intelligent analysis data.

[0004] A multi-task scheduling method for an artificial intelligence robot, the method comprising:

[0005] S1: Acquire a first facial image of a first user, and determine a first user portrait according to the first facial image;

[0006] The S1 comprises the following sub-steps:

[0007] S11: Obtain a first facial image of a first user, and determine whether the first user has a companion user according to the first behavior information of the first user, if yes, go to S12, otherwise go to S13;

[0008] S12: Acquire a first group image of a first group of people, where the first group of people includes the first user;

[0009] S13: Determine the first user portrait according to the first facial image and / or the first crowd image;

[0010] S2: determining first user activity information according to the first user portrait and a first user history portrait corresponding to the first user;

[0011] S3: determining a first reception robot according to the location information of the first user and the distribution information of the first robot;

[0012] The S3 comprises the following sub-steps:

[0013] S31: determining, according to the first robot distribution information, a plurality of first candidate robots whose distances from the first user are within a preset distance range;

[0014] S32: determining a first reception robot according to the first working state information of the first candidate robots;

[0015] S4: determining a first task to be performed according to the first user activity information and sending the task to the first reception robot;

[0016] S5: determining a second task to be performed according to the first feedback information of the first user and sending the second task to the second reception robot;

[0017] S6: Update the first user historical portrait of the first user.

[0018] Preferably, the S11 includes the following sub-steps:

[0019] S111: Acquire a first facial image of the first user;

[0020] S112: Acquire first behavior information of the first user, where the first behavior information includes a plurality of image frames;

[0021] S113: Inputting the first behavior information into a peer identification model to determine whether the first user has any peer users.

[0022] Preferably, the S13 includes the following sub-steps:

[0023] S130: If there is a user traveling with you, go to S131, otherwise go to S134;

[0024] S131: Determine first number of people information and a plurality of second facial images according to the first group image;

[0025] S132: re-identifying any user within a preset age range as a first user among the plurality of second facial images, and using the re-identified facial image of the first user as an updated first facial image;

[0026] S133: inputting the first number of people information, the updated first facial image, and a plurality of the second facial images into a first user portrait prediction model to output the first user portrait;

[0027] S134: Inputting the first facial image of the first user into a second user portrait prediction model to output the first user portrait.

[0028] Preferably, S2 comprises the following sub-steps:

[0029] S21: Determine whether the first user is a historical user, if yes, go to S22, otherwise go to S25;

[0030] S22: extracting a first user history portrait corresponding to the first user, calculating the similarity between the first user portrait and the first user history portrait, and if the similarity is greater than a preset value, going to S23, otherwise going to S24;

[0031] S23: Determine the first user activity information according to the first user history portrait, and go to S26;

[0032] S24: inputting the first user portrait and the first user history portrait into a first user behavior prediction model to determine first user activity information, and then proceeding to S26;

[0033] S25: Input the first user portrait into the second user behavior prediction model to determine the first user activity information, and then go to S26;

[0034] S26: End.

[0035] Preferably, S4 includes the following sub-steps:

[0036] S41: inputting the first user activity information and the first user portrait into a user activity prediction model, thereby determining A first tasks to be performed;

[0037] S42: determining a first activity position corresponding to the first task to be performed, and sending the first activity position and the first task to be performed to the first reception robot;

[0038] Wherein, A is a natural number greater than 0.

[0039] Preferably, S5 comprises the following sub-steps:

[0040] S51: the first reception robot sends each of the first tasks to be performed to a corresponding second reception robot according to the first robot distribution information and the first activity position;

[0041] S52: determining a second task to be performed according to first feedback information sent by the first user to the second reception robot;

[0042] S53: The cloud server sends the second task to be executed to the second reception robot.

[0043] The present application also proposes a multi-task scheduling system for an artificial intelligence robot, which is used to implement the multi-task scheduling method for the artificial intelligence robot, including:

[0044] A first user portrait acquisition module, used to acquire a first facial image of a first user, and determine a first user portrait according to the first facial image;

[0045] A first user activity information determination module, configured to determine first user activity information according to the first user portrait and a first user history portrait corresponding to the first user;

[0046] A first reception robot determining module, configured to determine a first reception robot according to the location information of the first user and the first robot distribution information;

[0047] A first task to be executed determining module, used for determining a first task to be executed according to the first user activity information and sending the first task to be executed to the first reception robot;

[0048] A second task to be performed determining module, used for determining a second task to be performed according to the first feedback information of the first user and sending the second task to be performed to the second reception robot;

[0049] The first user historical portrait updating module is used to update the first user historical portrait of the first user.

[0050] The present application proposes a multi-task scheduling method and device for an artificial intelligence robot. The first user's trip is divided into individual trips and group trips according to the trip scale. When the first user's matching degree with historical users meets the preset requirements, the first user's activity information is obtained according to the first user's historical portrait. The first user's image and the first group's image are combined to obtain the first user's portrait for the group's travel mode. The first user's portrait can better portray the group's travel preferences. When tasks are assigned, the robot's real-time location and the first user's feedback information are combined to update the tasks in real time, so that the robot's services can better meet the user's needs. The present application proposes a multi-task scheduling method and device for an artificial intelligence robot. The robot's services can better meet the user's actual needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0052] Figure 1 It is an execution flow chart of a multi-task scheduling method for an artificial intelligence robot of the present invention.

[0053] Figure 2 It is a system architecture diagram of the interaction between the cloud server and each artificial intelligence robot in the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments, wherein the illustrative embodiments and descriptions are only used to explain the present invention but are not intended to limit the present invention.

[0056] The following is a detailed description of a multi-task scheduling method and device for an artificial intelligence robot of the present invention.

[0057] This embodiment proposes a multi-task scheduling method for an artificial intelligence robot. The method flow is as follows: Figure 1 As shown, the specific steps include:

[0058] S1: Acquire a first facial image of a first user, and determine a first user portrait according to the first facial image.

[0059] When users enter spaces such as shopping malls, the artificial intelligence robots in the space need to predict the users' potential intentions as soon as possible, so as to provide them with targeted guidance and services.

[0060] S11: Acquire a first facial image of a first user, and determine whether the first user has any companions according to the first behavior information of the first user, if yes, go to S12, otherwise go to S13.

[0061] The S11 includes the following sub-steps:

[0062] S111: Acquire a first facial image of the first user.

[0063] The first facial image can be obtained through an image acquisition device such as a camera, and the first facial image is used for image recognition to match it with a historical user portrait.

[0064] S112: Acquire first behavior information of the first user, where the first behavior information includes a plurality of image frames.

[0065] The first behavior information includes whether there are other people within a preset distance range of the first user, and whether the first user has interactive behavior with the other people.

[0066] When it is detected that the first user has an interactive behavior with the other person, a number of image frames are captured by the camera.

[0067] S113: Inputting the first behavior information into a peer identification model to determine whether the first user has any peer users.

[0068] The peer discrimination model is obtained based on deep convolutional neural network (CNN) training, and the training samples include a number of image frame data for characterizing users traveling alone or in groups, and corresponding labeling data.

[0069] S12: Acquire a first group image of the first group.

[0070] The first group image includes overall image information of the first group and second image information of each person.

[0071] S13: Determine the first user portrait according to the first facial image and / or the first crowd image.

[0072] The first user portrait is used to characterize information such as the age, type, and behavioral preferences of the first user.

[0073] The S13 comprises the following sub-steps:

[0074] S130: If there is a fellow user, go to S131, otherwise go to S134.

[0075] S131: Determine first number of people information and a plurality of second facial images according to the first group image.

[0076] By analyzing the first crowd image, first number information included in the image can be obtained, and second facial images of other passers-by except the first user in the first crowd image can be obtained.

[0077] S132: Among the plurality of second facial images, re-identify any user within a preset age range as a first user, and use the re-identified facial image of the first user as an updated first facial image.

[0078] Usually, young and middle-aged users are the main decision-making users when a group of people travel. Therefore, in the case of a group of people traveling, users within a preset age range can be identified as the main predicted users, that is, the second users, and the facial image of the second user can be re-identified as the first facial image in S11.

[0079] S133: Input the first number of people information, the updated first facial image, and a plurality of the second facial images into a first user portrait prediction model to output the first user portrait.

[0080] The first user portrait prediction model is obtained based on deep convolutional neural network model training. In the training sample data, the number of people in the crowd, the first facial images of the main people and the second facial images of the passers-by are used as input information, and the type characteristic information of the crowd is used as output information.

[0081] Through the first user portrait prediction model, the personnel type composition of the crowd can be analyzed, such as the total number of people, age distribution, travel time and other information.

[0082] S134: Inputting the first facial image of the first user into a second user portrait prediction model to output the first user portrait.

[0083] If there is no peer user, then only the first facial image of the first user is input into the second user portrait prediction model.

[0084] The second user portrait prediction model is obtained based on deep convolutional neural network model training. In the training sample data, the first facial image is used as input information and the type characteristic information of the group of people is used as output information.

[0085] S2: Determine first user activity information according to the first user portrait and a first user history portrait corresponding to the first user.

[0086] S21: Determine whether the first user is a historical user, if so, go to S22, otherwise go to S25.

[0087] In the cloud server, the user's behavior information will be updated in real time according to the user's historical behavior. In this step, a traversal search is performed in the cloud server according to the first facial image to determine whether the first user is a historical user.

[0088] S22: extract the first user historical portrait corresponding to the first user, calculate the similarity between the first user portrait and the first user historical portrait, if the similarity is greater than a preset value, go to S23, otherwise go to S24.

[0089] For each historical user, the cloud server will analyze the behavior information and form the first user history portrait.

[0090] In this step, by calculating the similarity between the first user and the historical portrait of the first user, it can be determined whether the current travel of the first user has a high reproducibility with the historical travel behavior of the user.

[0091] For example, if the information recorded in the first user portrait includes that there are 4 people in this trip, including 2 elderly people, 1 child and the user, and the trip time is 11 a.m., and the first user's historical portrait also records travel behaviors similar to this trip, then relatively similar activity information can be determined for the first user based on the first user's historical travel behaviors. Otherwise, it is necessary to re-determine the recommended activity information for the first user based on the first user portrait corresponding to this trip.

[0092] S23: Determine the first user activity information according to the first user history portrait, and go to S26.

[0093] In this step, the activity information most relevant to the current trip in the historical portrait of the first user is determined as the activity information of the first user.

[0094] S24: Input the first user portrait and the first user history portrait into a first user behavior prediction model to determine first user activity information, and then go to S26.

[0095] The first user behavior prediction model is used when the first user profile has a low similarity with the first user history profile. For example, in the first user history profile, the user usually travels alone and the travel time is at night, while the first user profile this time includes companions and the travel time is noon, then the first user behavior prediction model needs to be used to re-predict the first user activity information.

[0096] During the training process of the first user behavior prediction model, the first user portrait and the first user historical portrait will be used as training samples at the same time, wherein the first user portrait provides information such as the distribution of people on this trip, which is used to determine the behavior type, and the first user historical portrait provides the user's historical preferences, which is used to further determine the specific user behavior under the determined behavior type.

[0097] The first user behavior prediction model is obtained based on deep convolutional neural network model training.

[0098] S25: Input the first user portrait into the second user behavior prediction model to determine the first user activity information, and then go to S26.

[0099] The second user behavior prediction model is used to make a preliminary prediction of the first user activity information based on the first user portrait.

[0100] The first user activity information may include shopping, buying clothes, eating, watching movies, etc.

[0101] S26: End.

[0102] S3: Determine a first reception robot according to the location information of the first user and the first robot distribution information.

[0103] In order to enable artificial intelligence robots to respond quickly to user needs, the robots need to be reasonably distributed in various locations.

[0104] The S3 comprises the following sub-steps:

[0105] S31: Determine, according to the first robot distribution information, a plurality of first candidate robots whose distances from the first user are within a preset distance range.

[0106] The first robot distribution information refers to the real-time distribution information of all artificial intelligence robots, which may specifically include coordinates.

[0107] S32: Determine a first reception robot according to the first working status information of the plurality of first candidate robots.

[0108] In this step, the working status information of the several candidate robots is obtained, and the first candidate robot whose idleness meets the specified requirements is determined as the first reception robot, so as to provide service for the first user.

[0109] S4: Determine a first task to be performed based on the first user activity information and send it to the first reception robot.

[0110] S41: Inputting the first user activity information and the first user portrait into a user activity prediction model to determine A first tasks to be performed.

[0111] In this step, the input first user activity information is used to limit the activity category, and the input first user portrait is used to further find a plurality of first tasks to be performed that meet the first user's preferences under the activity category.

[0112] For example, if the first activity information is eating, then the plurality of first tasks to be performed may be a number of restaurants that match the first user's preferences.

[0113] S42: Determine a first activity position corresponding to the first task to be performed, and send the first activity position and the first task to be performed to the first reception robot.

[0114] Each of the first tasks to be performed corresponds to a first activity position.

[0115] S5: Determine a second task to be performed according to the first feedback information of the first user and send it to the second reception robot.

[0116] S51: The first reception robot sends each of the first tasks to be performed to a corresponding second reception robot according to the first robot distribution information and the first activity position.

[0117] In this step, based on the first activity position corresponding to each of the first tasks to be performed, several robots within a preset distance from the first activity position are selected as alternative robots, and the second reception robot corresponding to each of the first tasks to be performed is determined in combination with the real-time working status information of each alternative robot.

[0118] S52: Determine a second task to be executed according to the first feedback information sent by the first user to the second reception robot.

[0119] When the second reception robot provides services to the first user, the first user may give first feedback information based on the satisfaction with the service results, and the second reception robot may determine a second task to be performed based on the first feedback information.

[0120] For example, when the second reception robot provides a guide service for the first user, the first user may suddenly need to go to the bathroom, so going to the bathroom is the first feedback information.

[0121] S53: The cloud server sends the second task to be executed to the second reception robot.

[0122] After receiving the first feedback information, the second reception robot can upload the first feedback information to the cloud server, and the cloud server will perform semantic analysis on the first feedback information, and combine it with the location information to generate the second task to be executed and send it to the second reception robot.

[0123] S6: Update the first user historical portrait of the first user.

[0124] The first user historical portrait is used to describe the behavioral preference information of the first user and needs to be updated in real time. Therefore, a new label is added to the first user historical portrait according to the current user behavior, and the old label is updated.

[0125] For example, if this is the first time that the first user travels in a group, labels such as group travel and group travel preference need to be added to the first user's historical portrait.

[0126] The system architecture diagram of the cloud server and each artificial intelligence robot is as follows: Figure 2As shown, basic computing problems are completed by each artificial intelligence robot, and complex computing problems are completed by a cloud server, which communicates remotely with each artificial robot. This application also proposes a multi-task scheduling system for an artificial intelligence robot, which is used to implement the multi-task scheduling method of the artificial intelligence robot, including:

[0127] A first user portrait acquisition module, used to acquire a first facial image of a first user, and determine a first user portrait according to the first facial image;

[0128] A first user activity information determination module, configured to determine first user activity information according to the first user portrait and a first user history portrait corresponding to the first user;

[0129] A first reception robot determining module, configured to determine a first reception robot according to the location information of the first user and the first robot distribution information;

[0130] A first task to be executed determining module, used for determining a first task to be executed according to the first user activity information and sending the first task to be executed to the first reception robot;

[0131] A second task to be performed determining module, used for determining a second task to be performed according to the first feedback information of the first user and sending the second task to be performed to the second reception robot;

[0132] The first user historical portrait updating module is used to update the first user historical portrait of the first user.

[0133] The present application proposes a multi-task scheduling method and device for an artificial intelligence robot. The first user's trip is divided into individual trips and group trips according to the trip scale. When the first user's matching degree with historical users meets the preset requirements, the first user's activity information is obtained according to the first user's historical portrait. The first user's image and the first group's image are combined to obtain the first user's portrait for the group's travel mode. The first user's portrait can better portray the group's travel preferences. When tasks are assigned, the robot's real-time location and the first user's feedback information are combined to update the tasks in real time, so that the robot's services can better meet the user's needs. The present application proposes a multi-task scheduling method and device for an artificial intelligence robot. The robot's services can better meet the user's actual needs.

[0134] The above description is only a preferred embodiment of the present invention, so all equivalent changes or modifications made according to the structure, characteristics and principles described in the scope of the patent application of the present invention are included in the scope of the patent application of the present invention.

Claims

1. A multi-task scheduling method for an artificial intelligence robot, characterized in that: The method includes: S1: Acquire a first facial image of a first user, and determine a first user portrait according to the first facial image; The S1 comprises the following sub-steps: S11: Obtain a first facial image of a first user, and determine whether the first user has a companion user according to the first behavior information of the first user, if yes, go to S12, otherwise go to S13; S12: Acquire a first group image of a first group of people, where the first group of people includes the first user; S13: Determine the first user portrait according to the first facial image and / or the first crowd image; S2: determining first user activity information according to the first user portrait and a first user history portrait corresponding to the first user; S3: determining a first reception robot according to the location information of the first user and the distribution information of the first robot; The S3 comprises the following sub-steps: S31: determining, according to the first robot distribution information, a plurality of first candidate robots whose distances from the first user are within a preset distance range; S32: determining a first reception robot according to the first working state information of the first candidate robots; S4: determining a first task to be performed according to the first user activity information and sending the task to the first reception robot; S5: determining a second task to be performed according to the first feedback information of the first user and sending the second task to the second reception robot; S6: updating the first user history portrait of the first user; S11 includes the following sub-steps: S111: Acquire a first facial image of the first user; S112: Acquire first behavior information of the first user, where the first behavior information includes a plurality of image frames; S113: inputting the first behavior information into a peer identification model to determine whether the first user has a peer user; S13 includes the following sub-steps: S130: If there is a user traveling with you, go to S131, otherwise go to S134; S131: Determine first number of people information and a plurality of second facial images according to the first group image; S132: re-identifying any user within a preset age range as a first user among the plurality of second facial images, and using the re-identified facial image of the first user as an updated first facial image; S133: inputting the first number of people information, the updated first facial image, and a plurality of the second facial images into a first user portrait prediction model to output the first user portrait; S134: Inputting the first facial image of the first user into a second user portrait prediction model to output the first user portrait.

2. The multi-task scheduling method of an artificial intelligence robot according to claim 1, characterized in that: The S2 comprises the following sub-steps: S21: Determine whether the first user is a historical user, if yes, go to S22, otherwise go to S25; S22: extracting a first user history portrait corresponding to the first user, calculating the similarity between the first user portrait and the first user history portrait, and if the similarity is greater than a preset value, going to S23, otherwise going to S24; S23: Determine the first user activity information according to the first user history portrait, and go to S26; S24: inputting the first user portrait and the first user history portrait into a first user behavior prediction model to determine first user activity information, and then proceeding to S26; S25: Input the first user portrait into the second user behavior prediction model to determine the first user activity information, and then go to S26; S26: End.

3. The multi-task scheduling method of an artificial intelligence robot according to claim 2, characterized in that: The S4 comprises the following sub-steps: S41: inputting the first user activity information and the first user portrait into a user activity prediction model, thereby determining A first tasks to be performed; S42: determining a first activity position corresponding to the first task to be performed, and sending the first activity position and the first task to be performed to the first reception robot; Wherein, A is a natural number greater than 0.

4. The multi-task scheduling method of an artificial intelligence robot according to claim 3, characterized in that: The S5 comprises the following sub-steps: S51: the first reception robot sends each of the first tasks to be performed to a corresponding second reception robot according to the first robot distribution information and the first activity position; S52: determining a second task to be performed according to first feedback information sent by the first user to the second reception robot; S53: The cloud server sends the second task to be executed to the second reception robot.

5. A multi-task scheduling system for an artificial intelligence robot, used to implement a multi-task scheduling method for an artificial intelligence robot in claim 1, comprising: A first user portrait acquisition module, used to acquire a first facial image of a first user, and determine a first user portrait according to the first facial image; A first user activity information determination module, configured to determine first user activity information according to the first user portrait and a first user history portrait corresponding to the first user; A first reception robot determining module, configured to determine a first reception robot according to the location information of the first user and the first robot distribution information; A first task to be executed determining module, used for determining a first task to be executed according to the first user activity information and sending the first task to be executed to the first reception robot; A second task to be performed determining module, used for determining a second task to be performed according to the first feedback information of the first user and sending the second task to be performed to the second reception robot; The first user historical portrait updating module is used to update the first user historical portrait of the first user.

Images