CN110866588B - Training learning method and system for realizing individuation of learning ability model of intelligent virtual digital animal - Google Patents
Training learning method and system for realizing individuation of learning ability model of intelligent virtual digital animal Download PDFInfo
- Publication number
- CN110866588B CN110866588B CN201911086750.4A CN201911086750A CN110866588B CN 110866588 B CN110866588 B CN 110866588B CN 201911086750 A CN201911086750 A CN 201911086750A CN 110866588 B CN110866588 B CN 110866588B
- Authority
- CN
- China
- Prior art keywords
- animal
- virtual digital
- learning ability
- intelligent virtual
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/008—Artificial life, i.e. computing arrangements simulating life based on physical entities controlled by simulated intelligence so as to replicate intelligent life forms, e.g. based on robots replicating pets or humans in their appearance or behaviour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/60—Software deployment
- G06F8/65—Updates
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Robotics (AREA)
- Computer Security & Cryptography (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a training learning method and a training learning system for realizing individuation of a learning ability model of an intelligent virtual digital animal. The method comprises the following steps: generating a program instance of the intelligent virtual digital animal at the cloud; constructing a real object map of the production animal robot for the intelligent virtual digital animal cloud instance, and distributing the real object map to a user, and downloading a corresponding learning ability model from the cloud by the animal robot; the user and the animal robot real object perform natural scene interaction, the animal robot automatically collects, generates personalized training data, trains and updates the copy of the learning ability model and uploads the copy to the cloud, and the cloud updates the learning ability model of the corresponding virtual digital animal. The invention effectively solves the problems that the existing artificial intelligent model training mode has too high threshold for common people, cannot exert the value of mass labor force in a large range, cannot participate in the personalized training of the virtual digital intelligent capability model, and lacks the real scene of user interaction.
Description
Technical Field
The invention belongs to the technical field of machine learning training, and particularly relates to a training learning method and system for realizing individuation of a learning ability model of an intelligent virtual digital animal.
Background
The virtual digital animal is a virtual object characterized by a basic attribute list and a behavior list through a computer program data structure and a code segment at a cloud server side, wherein the virtual object is usually a conceptual mapping of one or more real animals in a real scene, and the conceptual mapping can even overlay the attributes of other biological species and fictive species. With the development of artificial intelligence, it has been proposed to add a "learning ability model" intelligent module, called an intelligent virtual digital animal, to the virtual digital animal that can be learned by training. The "learning ability model" intelligent module referred to herein can be typically represented by constructing a neural network model, while training learning supporting multiple abilities can be achieved by designing multiple intelligent modules for intelligent virtual digital animals that are oriented toward learning of different abilities.
In recent years, with the rapid progress of artificial intelligence technology theory and method and the great progress of intelligent manufacturing capability (the layers of intelligent agents such as machine dogs, robots, robot bee groups and the like are endless, updated), users bring forward personalized training learning capability and interactive mode innovation (traditional virtual scene interaction completely in a software interface, such as using a mouse or a touch screen, and interaction with a virtual pet through virtual graphics and picture interfaces drawn by software) on traditional intelligent virtual digital animals with pure software interaction.
The current intelligent virtual digital animal training intelligent module is technically characterized in that professional computer talents train a learnable model by professional means such as programming and the like through data collected in a unified way in advance to obtain a unified capacity model, and the obtained unified capacity model is copied to user program object examples of intelligent virtual digital animals in an indiscriminate manner. The method or the mode of the training technology has the defects and the shortcomings which are summarized as 'lack of personalized user real interaction scene data for personalized learning training of intelligent virtual digital animals', 'lack of physical interaction between a person in a real scene and the virtual digital animals', and the method and the device are specifically expressed in the following steps: the user (ordinary person) cannot train and teach own intelligent virtual digital animals individually and directly, and can only interact with own intelligent virtual digital animals in a software scene, but cannot interact with own intelligent virtual digital animals in a physical world truly. Therefore, it is difficult to satisfy the higher demands for personalized training learning ability such as intelligent digital pet dogs partially or completely replacing biological dogs (owners can selectively teach personalized skills of biological dog society according to own needs), enabling common people to participate in intelligent virtual digital animal training as domesticated animals without threshold, and trainable learning growth intelligent service robot dogs, intelligent service robots and the like aiming at specific user behavior habits.
Disclosure of Invention
The invention aims at: the method overcomes the defects of the prior training learning technology in a method mode, and provides a novel training learning method and system for realizing individuation of a learning ability model of an intelligent virtual digital animal.
The invention manufactures a corresponding animal robot for each intelligent virtual digital animal, and the animal robot synchronously downloads a learning ability model intelligent module of the intelligent virtual digital animal from a cloud and stores the learning ability model intelligent module as a local copy. In the real scene that each user (ordinary person) directly interacts with the animal robot belonging to the user, or in the process that the user consciously trains (teaching, note: different from the traditional configuration of some fixed parameters) the user's own animal robot, personalized training data is automatically generated, a copy of a learning-capable model is trained, model parameters are updated, and the model parameters are synchronized to the cloud, so that the problems of ' lack of personalized learning training of intelligent virtual digital animals by personalized user real interaction scene data ', ' lack of physical interaction between the real scene person and the virtual digital animals ', and the like are solved.
The technical scheme adopted by the invention is as follows:
in a first aspect, the present invention provides a training learning method for implementing individualization of a learning ability model of an intelligent virtual digital animal, comprising the steps of:
downloading a copy of the learning ability model from the cloud server by the animal robot entity; the animal robot physical object is an animal robot manufactured according to an intelligent virtual digital animal example of a cloud server side;
the animal robot physical object realizes personalized training of the learning ability model of the intelligent virtual digital animal by carrying out scene interaction with a user;
and uploading the trained learning ability model copy to the cloud server by the animal robot entity so that the cloud server can update the learning ability model of the intelligent virtual digital animal according to the uploaded learning ability model copy.
Further, the animal robot entity downloads a copy of the learning ability model from the cloud server, including:
the animal robot entity sends an identity verification request to the cloud server end, so that the cloud end can carry out identity verification after receiving the request, and write and lock a corresponding learning ability model after the verification is passed;
the animal robot entity requests to download the learning ability model to the cloud server;
verifying the data integrity of the downloaded learnable capacity model by the animal robot real object;
the animal robot physical object loads the downloaded learning ability model and sets the learning ability model initial copy of the animal robot physical object.
Further, the animal robot physical object realizes personalized training of the learning ability model of the intelligent virtual digital animal by performing scene interaction with a user, and comprises the following steps:
the user selects a training target for the animal robot;
the user performs personalized scene interaction with the animal robot real object;
in the interaction process, the animal robot real object automatically collects and generates personalized training data through a sensor, and marks the data according to the training target;
based on the collected data, the automatic training updates the learning ability model replica.
In a second aspect, the present invention provides a training learning method for implementing individualization of a learning ability model of an intelligent virtual digital animal, comprising the steps of:
the cloud server side generates an intelligent virtual digital animal instance;
the cloud server side sends a copy of the learning ability model to the animal robot real object according to the request of the animal robot real object; the animal robot physical object is an animal robot manufactured according to an intelligent virtual digital animal example of a cloud server side;
the cloud server receives the copy of the learning ability model which is uploaded by the animal robot and is trained, and updates the learning ability model of the intelligent virtual digital animal; the learning ability model copy after training is completed is the learning ability model copy after the user performs scene interaction with the animal robot real object and personalized training.
Further, the cloud server side generates an intelligent virtual digital animal instance by adopting the following steps:
registering the user;
the user selects the type of intelligent virtual digital animal to be picked up;
an intelligent virtual digital animal program instance is generated for a user.
Further, the updating the learning ability model of the intelligent virtual digital animal comprises:
verifying the data integrity of the copy of the learning ability model;
releasing the learning ability model write lock;
and updating the learning ability model of the intelligent virtual digital animal according to the uploaded learning ability model copy.
Further, the cloud server side is provided with an intelligent virtual digital animal program object instance cloud hosting platform, the intelligent virtual digital animal program object instance cloud hosting platform runs program logic of the intelligent virtual digital animal at a cloud end, stores user resources, stores and manages virtual digital animal types, all attributes and behavior lists of the intelligent virtual digital animal, and attribute lists, is responsible for interacting with animal robot objects, and provides access pages for users to directly access; the virtual digital animal type defines data structures of different animal categories, including animal attributes, animal behaviors, frames of animal learning ability models; the intelligent virtual digital animal is an instantiation program object of an intelligent virtual digital animal type and consists of a learning ability model of the intelligent virtual digital animal, an attribute list and a behavior list, wherein the attribute list records attribute values of an intelligent virtual animal instance, and the behavior list defines basic action behaviors of the animal.
Further, the learning ability model is composed of a learning ability model structure definition, model parameters and model parameter integrity check values; and obtaining a group of parameter sets, namely model parameters, through training and learning, and generating an integrity check value for each group of stable model parameters.
In a third aspect, the present invention provides an animal robot comprising a physical body and a computing system; the computing system downloads a learning ability model copy of the intelligent virtual digital animal from the cloud server, realizes personalized training of the learning ability model by performing scene interaction with a user, and uploads the trained learning ability model copy to the cloud server, so that the cloud server can update the learning ability model of the intelligent virtual digital animal according to the uploaded learning ability model copy.
Further, the animal robot is manufactured according to the structural body information of the intelligent virtual digital animal instance at the cloud server end, is a physical mapping of the intelligent virtual digital animal instance, and realizes the extension of the virtual digital world to the physical world; the physical machine body comprises physical presentation of attributes and behaviors of the intelligent virtual digital animal, wherein the physical presentation comprises colors, textures, a programmable control manipulator, a mechanical arm and a visual sensor, and the computing system comprises an example management agency module, a learning ability model copy module, an NPU module, a control logic integrated circuit and a storage module.
Further, the instance management agent module communicates with an instance management module at the cloud server end through a network and is responsible for submitting an identity authentication request to the cloud hosting platform, downloading a learning ability model, checking the downloaded learning ability model, installing the learning ability model as a local model copy of an animal robot real object, scheduling a program for training and updating the learning ability model copy, and uploading the training and updated learning ability model copy to the cloud; the NPU module is a special intelligent computing processor for accelerating the reasoning and training performance of the neural network, the control logic integrated circuit is a processor for realizing programmable control of the animal robot, and the storage module provides a rapid storage function for the copy of the learning ability model, training data and program operation.
In a fourth aspect, the invention provides a cloud server provided with an intelligent virtual digital animal program object instance cloud hosting platform;
the intelligent virtual digital animal program object instance cloud hosting platform generates an intelligent virtual digital animal instance, and sends a learnable capacity model copy to an animal robot object according to a request of the animal robot object; the animal robot physical object is an animal robot manufactured according to an intelligent virtual digital animal example of a cloud server side;
the intelligent virtual digital animal program object instance cloud hosting platform receives the training-completed learning ability model copy uploaded by the animal robot object and updates the learning ability model of the intelligent virtual digital animal; the learning ability model copy after training is completed is the learning ability model copy after the user performs scene interaction with the animal robot real object and personalized training.
Further, the cloud server comprises an instance management module, a virtual digital animal type module and an intelligent virtual animal instance module;
the instance management module manages the full life cycle of the intelligent virtual digital animal instance and the learning ability model management of the instance and is responsible for interactive communication with the instance management agent module running in the animal robot real object through a network;
the virtual digital animal type module comprises different animal type, and defines data structures of different animal types, including animal attributes, animal behaviors and frames of animal learning ability models;
the intelligent virtual animal instance module is a specific instantiation of intelligent virtual digital animal types and comprises a learning ability model, an attribute list and a behavior list.
In a fifth aspect, the present invention provides a training learning system for implementing individualization of a learning ability model of an intelligent virtual digital animal, which includes the above-described animal robot and a cloud server.
Compared with the prior art, the invention has the advantages that:
(1) Currently, training of artificial intelligence capability models (such as RNN, CNN networks, deep reinforcement models, etc.) of intelligent virtual digital animals requires specialized computer engineers, with high threshold; the invention provides a method and a mode for training a capacity model for an intelligent virtual digital animal in natural scene interaction of an ordinary person with an off-line animal robot object for mapping the off-line animal robot object of an on-line intelligent virtual digital animal production line, which obviously reduces a threshold of an artificial intelligent capacity model for training the virtual digital animal.
(2) At present, the learning ability model training of the intelligent virtual digital animal adopts the modes of unified data collection, unified training and unified deployment before online or online, and has the problems of lack of personalized intelligence, flexibility and real scene participation; the training learning method and system provided by the invention are very flexible, and solve the problems of individuation of the learning ability model of the intelligent virtual digital animal for a large number of users, physical interaction of the human and the real scene of the virtual digital animal, and the like. The training and interaction mode of combining the online intelligent virtual digital example and the offline physical robot mapping provides a technical method and mode for realizing the application of digital pets and the like with the continuous ' learning growth ' of personalized capabilities (such as learning to call, recognize colors, pick up thrown balls and the like) under the ' training of a master.
Drawings
FIG. 1 is a schematic view of the overall structure of the present invention;
FIG. 2 is a process of generating an animal robot map for intelligent virtual digital animal construction in accordance with the present invention;
fig. 3 is a training learning process of the present invention that implements the personalisation of the learning ability model of the intelligent virtual digital animal.
Detailed Description
The invention is further illustrated in the following figures and examples, which are not intended to limit the scope of the invention in any way.
The training learning system for realizing individuation of the learning ability model of the intelligent virtual digital animal in the embodiment, as shown in fig. 1, mainly comprises an intelligent virtual digital animal program object instance cloud hosting platform 1, an animal robot physical object 2 and a network 3. The intelligent virtual digital animal program object instance cloud hosting platform 1 mainly comprises an instance management module 101, a virtual digital animal type module 102, an intelligent virtual digital animal instance module 103 (a plurality of different instances form an instance set) and the like. Intelligent virtual digital animal instantiation module 103 contains a learning ability model 1031, a list of attributes 1032, and a list of behaviors 1033. The animal robot physical object 2 is composed of an example management agency module 201, a learning ability model copy module 202, an NPU (neural network processor) module 203, a storage module 205, a control logic integrated circuit 204, a physical body and the like. The modules represented by dashed boxes in the figures are not a central component of the invention. The person 4 is a user, is an owner of a certain intelligent virtual digital animal example and a corresponding animal robot real object, directly interacts with the corresponding animal robot real object in a physical scene, and the animal robot real object collects and generates training data in the interaction through a visual sensor, a collision detection sensor, a touch sensor, a pressure sensor, an optical fiber sensor and the like, trains and updates a copy of the learning ability model. The cloud hosting platform 1 of the intelligent virtual digital animal program object instance runs on the cloud, the animal robot object 2 is at a user end (an ordinary person), and the cloud and the animal robot object are connected through the network 3 to provide a data transmission link.
The cloud hosting platform 1 of the intelligent virtual digital animal program object instance runs intelligent virtual digital animal program logic at the cloud, stores user resources, stores and manages virtual digital animal types, all attributes and behavior attribute lists of the intelligent virtual digital animal, is responsible for interacting with animal robot objects, and also provides access pages for users (people) to directly access and the like. Wherein the instance management module 101 manages the full life cycle of the intelligent virtual digital animal instance, the learning ability model management of the instance (locking the learning ability model, unlocking the learning ability model, updating the learning ability model, etc.), and is responsible for interactive communication with the instance management agent module 201 running in the animal robot physical object 2 through the network 3, the communication content mainly comprises verifying the identity of the user and the instance, providing the learning ability model download, receiving the learning ability model upload, checking the learning ability model, etc.
The virtual digital animal type module 102 is a framework for platform-managed different animal class types, defining data structures of different animal classes, such as animal attributes, animal behaviors, and animal learning ability models.
The intelligent virtual animal instance module 103 is a specific instantiation of intelligent virtual digital animal types, each animal type typically has multiple instances, each instance has an independently stored learning ability model, and all the instances constitute an instance set. The smart virtual animal instance is composed of a learning ability model 1031, a list of attributes 1032, a list of behaviors 1033, and the like. Wherein the attribute list defines attributes (such as unique Identification (ID), color, age, gender, etc.) of the intelligent virtual animal instance, and the behavior list defines basic action behaviors (such as running, walking, turning, jumping, vigorous, etc.) of the animal.
The learning ability model 1031 is composed of a learning model structure definition, model parameters and model parameter integrity check values, wherein the learning model structure can be an Artificial Neural Network (ANN), deep reinforcement learning (Reinforcement Learning), a pulse neural network (SNN) and the like, and different types of the learning model structure do not affect the rights of the invention. A set of parameters, i.e. model parameters, is obtained by training and learning, and for each set of stable model parameters, a hash value is generated as an integrity check value (the integrity check value generation algorithm is an option, and selecting other algorithms than the hash algorithm does not affect the rights of the invention).
The animal robot physical object 2 is an animal robot manufactured according to the structural body information of the intelligent virtual digital animal example module 103, and is a physical mapping of the intelligent virtual digital animal example, so that the extension of the virtual digital world to the physical world is realized. The method and the mode for training the intelligent virtual digital animal are provided for the first time, and are the key step for realizing that common mass users (instead of professional computer talents) can purposefully generate data and a training model in combination with personal wish in a real interactive scene, thereby achieving the individualized capability training of the intelligent virtual digital animal. The animal robot consists of physical body and calculating system. The physical machine body comprises physical presentation of the attribute and the behavior of the intelligent virtual digital animal, such as color, texture, a programmable control manipulator, a mechanical arm, a visual sensor and the like; the design, implementation, manufacturing method of these physical machine bodies are not important components of the present invention, and different types of manufacturing methods, technologies, processes, etc. do not affect the rights of the present invention. The computing system is mainly composed of an instance management agent module 201, a learning ability model copy module 202, an NPU module 203, a control logic integrated circuit 204 and a storage module 205.
The instance management agent module 201 communicates with the instance management module 101 through the network 3, and is responsible for submitting an identity authentication request to the cloud hosting platform, downloading the learning ability model 1031, checking the downloaded learning ability model 1031, installing the learning ability model 1031 as a local copy, and uploading the updated learning ability model copy 202 to the cloud. The NPU module 203 is a special intelligent computing processor for accelerating neural network reasoning and training performance, the control logic integrated circuit 204 is a processor for realizing programmable control of the animal robot, the storage module 205 provides fast storage for copies of the learning ability model, training data, program running and the like, and different types of the modules 203, 204 and 205 do not affect the rights of the invention.
The invention is further described below with reference to fig. 2 and 3 by two general users, each training their own intelligent virtual digital dog to learn differently skills. Fig. 2 illustrates a flow for generating an animal robot physical map for intelligent virtual digital animal construction, and fig. 3 illustrates a training learning flow for implementing individualization of a learning ability model of an intelligent virtual digital animal.
(1) An animal robot physical map is generated for the intelligent virtual digital animal construct. As shown in fig. 2, the method mainly comprises the following steps:
(1.1) registering an intelligent virtual digital animal program object instance cloud hosting platform 1 by a user A, and acquiring an intelligent virtual digital animal instance; in this embodiment, the user first gets an intelligent virtual digital dog whose attribute list includes unique identity number 10001, color gray, age 3 months, etc.; the learning ability model of the 10001 intelligent virtual digital dog integrates model frames of image recognition, voice understanding, motion trail tracking and the like based on deep learning, so that the learning ability model has the capabilities of obtaining a recognition image, understanding voice, automatically adjusting motion trail and the like through training learning. Wherein the image recognition model is realized based on a CNN network (convolutional neural network);
(1.2) producing an animal robot physical map for manufacturing a virtual digital dog number 10001;
(1.3) assigning 10001 number animal robot real objects to user A;
(1.4) starting 10001 number animal robot real object by the user A, and networking the robot;
(1.5) the 10001 number animal robot entity sends an identity verification request to the cloud hosting platform;
(1.6) the cloud instance management module receives the request, performs identity verification, identifies the animal robot entity as the animal robot entity 10001 (if the identity verification fails, the animal robot entity feeds back identity verification failure information to the animal robot entity, and the animal robot entity can resend the identity verification request), and then performs write locking on the learning ability model of the intelligent virtual digital animal 10001 (avoids inconsistent models or write conflicts caused by parallel writing);
(1.7) a 10001 animal robot physical request downloads a learning ability model of a 10001 intelligent virtual digital animal hosted by a cloud, the downloading is completed and the model integrity check is passed (if the model is not passed, the model is not completely downloaded and needs to be downloaded again);
and (1.8) loading the downloaded model, and initializing the model as a learning ability model copy of the 10001-model animal robot.
(2) And performing personalized training on the learning ability model of the intelligent virtual digital animal. As shown in fig. 3, the method mainly comprises the following steps:
(2.1) user A selects a training target (which can be realized by techniques such as voice awakening and menu selection and the like, and the mode is not limited) through a 10001 animal robot object: recognizing apples by a school; then the robot physical object enters a training interaction mode;
(2.2) displaying different apples and different sides of the apples in front of the 10001 animal robot, wherein the animal robot acquires the pictures of the apples through a vision sensor, and each movement and angle transformation of the apples triggers the acquisition of new pictures;
(2.3) in combination with the training targets set in step 2.1), all the collected photos are labeled as "apples", the 10001 animal robot accelerates by using the NPU, and the image classification subsystem model of the copy of the learning ability model is trained and updated until convergence. The image classification subsystem is realized based on the CNN convolutional neural network, the multi-classifier and other technologies (but other technical choices do not affect the rights of the invention);
(2.4) uploading the training updated learning ability model copy to the cloud end by the 10001 animal robot;
(2.5) the cloud instance management module receives the uploading and verifies the data integrity of the model (if the integrity verification is not passed, the model uploading is incomplete and needs to be uploaded again);
(2.6) the cloud instance management module releases the write lock of the learning ability model of the 10001 intelligent virtual digital animal instance;
and (2.7) updating the learnable capacity model of the 10001-number virtual digital animal by the cloud end according to the uploaded capacity model copy. The intelligent virtual digital animal of the user A then has the capability of individually identifying apples. In the training process, the user A generates personalized training data through the scene interaction help model, and is suitable for the common masses.
The unique identity number of the intelligent virtual digital dog which is picked up by the user B is 10002, the user B can train the intelligent virtual digital dog of the user B to obtain personalized ability for identifying football, basketball, orange and the like through a process similar to that of the user A, and a common user can train the intelligent virtual digital animal of the user B to obtain track tracking ability and other learning ability of a specific form by adopting a similar training method mode. Under the low-technical-threshold training method and mode, the intelligent virtual digital dogs of the user A and the user B show differentiated and personalized trainable learning ability 'growth'.
The invention provides a training learning method and a training learning system for realizing individuation of a learning ability model of an intelligent virtual digital animal, which provides a simple, convenient and efficient way for common masses to participate in the ability model training of a virtual digital intelligent body.
In the specific implementation of the scheme of the invention, after a user helps an animal robot entity to generate or collect personalized training data through scene interaction in the training process, if a system does not immediately perform model training at the animal robot entity end, the data is firstly uploaded to the cloud end and is deferred until the learning ability model of the corresponding virtual digital animal of the user is updated in the cloud end training, thereby realizing individuation of the learning ability model, and the method is regarded as a deformation mode of the invention.
The above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and those skilled in the art may modify or substitute the technical solution of the present invention without departing from the principle and scope of the present invention, and the protection scope of the present invention shall be defined by the claims.
Claims (12)
1. A training learning method for realizing individualization of a learning ability model of an intelligent virtual digital animal comprises the following steps:
downloading a copy of the learning ability model from the cloud server by the animal robot entity; the animal robot physical object is an animal robot manufactured according to an intelligent virtual digital animal example of a cloud server side;
the animal robot physical object realizes personalized training of the learning ability model of the intelligent virtual digital animal by carrying out scene interaction with a user;
uploading the trained learning ability model copy to the cloud server by the animal robot entity so that the cloud server can update the learning ability model of the intelligent virtual digital animal according to the uploaded learning ability model copy;
the animal robot is manufactured according to the structural body information of the intelligent virtual digital animal instance at the cloud server end, is used for physically mapping the intelligent virtual digital animal instance, and realizes the extension of the virtual digital world to the physical world; the animal robot includes a physical body and a computing system; the computing system downloads a learning ability model copy of the intelligent virtual digital animal from the cloud server, realizes personalized training of the learning ability model by performing scene interaction with a user, and uploads the trained learning ability model copy to the cloud server, so that the cloud server can update the learning ability model of the intelligent virtual digital animal according to the uploaded learning ability model copy; the physical body comprises a physical presentation of attributes and behaviors of the intelligent virtual digital animal;
the cloud server end is provided with an intelligent virtual digital animal program object instance cloud hosting platform, the intelligent virtual digital animal program object instance cloud hosting platform runs program logic of an intelligent virtual digital animal at a cloud end, stores user resources, stores and manages virtual digital animal types, all attributes, behavior lists and attribute lists of the intelligent virtual digital animal, is responsible for interacting with animal robot objects, and provides access pages for users to directly access; the virtual digital animal type defines data structures of different animal categories, including animal attributes, animal behaviors, and frames of animal learning ability models; the intelligent virtual digital animal is an instantiation program object of an intelligent virtual digital animal type and consists of a learning ability model of the intelligent virtual digital animal, an attribute list and a behavior list, wherein the attribute list records attribute values of an intelligent virtual animal instance, and the behavior list defines basic action behaviors of the animal.
2. The method of claim 1, wherein the animal robot entity downloads a copy of the learning ability model from a cloud server, comprising:
the animal robot entity sends an identity verification request to the cloud server end, so that the cloud end can carry out identity verification after receiving the request, and write and lock a corresponding learning ability model after the verification is passed;
the animal robot entity requests to download the learning ability model to the cloud server;
verifying the data integrity of the downloaded learnable capacity model by the animal robot real object;
the animal robot physical object loads the downloaded learning ability model and sets the learning ability model initial copy of the animal robot physical object.
3. The method of claim 1, wherein the animal robot entity performs the personalized training of the learning ability model of the intelligent virtual digital animal by performing a scene interaction with the user, comprising:
the user selects a training target for the animal robot;
the user performs personalized scene interaction with the animal robot real object;
in the interaction process, the animal robot real object automatically collects and generates personalized training data through a sensor, and marks the data according to the training target;
based on the collected data, the automatic training updates the learning ability model replica.
4. A training learning method for realizing individualization of a learning ability model of an intelligent virtual digital animal comprises the following steps:
the cloud server side generates an intelligent virtual digital animal instance;
the cloud server side sends a copy of the learning ability model to the animal robot real object according to the request of the animal robot real object; the animal robot physical object is an animal robot manufactured according to an intelligent virtual digital animal example of a cloud server side;
the cloud server receives the copy of the learning ability model which is uploaded by the animal robot and is trained, and updates the learning ability model of the intelligent virtual digital animal; the learning ability model copy after training is completed is a learning ability model copy after the user performs scene interaction with the animal robot real object and personalized training;
the animal robot is manufactured according to the structural body information of the intelligent virtual digital animal instance at the cloud server end, is used for physically mapping the intelligent virtual digital animal instance, and realizes the extension of the virtual digital world to the physical world; the animal robot includes a physical body and a computing system; the computing system downloads a learning ability model copy of the intelligent virtual digital animal from the cloud server, realizes personalized training of the learning ability model by performing scene interaction with a user, and uploads the trained learning ability model copy to the cloud server, so that the cloud server can update the learning ability model of the intelligent virtual digital animal according to the uploaded learning ability model copy; the physical body comprises a physical presentation of attributes and behaviors of the intelligent virtual digital animal;
the cloud server end is provided with an intelligent virtual digital animal program object instance cloud hosting platform, the intelligent virtual digital animal program object instance cloud hosting platform runs program logic of an intelligent virtual digital animal at a cloud end, stores user resources, stores and manages virtual digital animal types, all attributes, behavior lists and attribute lists of the intelligent virtual digital animal, is responsible for interacting with animal robot objects, and provides access pages for users to directly access; the virtual digital animal type defines data structures of different animal categories, including animal attributes, animal behaviors, and frames of animal learning ability models; the intelligent virtual digital animal is an instantiation program object of an intelligent virtual digital animal type and consists of a learning ability model of the intelligent virtual digital animal, an attribute list and a behavior list, wherein the attribute list records attribute values of an intelligent virtual animal instance, and the behavior list defines basic action behaviors of the animal.
5. The method of claim 4, wherein the cloud server side generates the intelligent virtual digital animal instance by:
registering the user;
the user selects the type of intelligent virtual digital animal to be picked up;
an intelligent virtual digital animal program instance is generated for a user.
6. The method of claim 4, wherein updating the learning ability model of the intelligent virtual digital animal comprises:
verifying the data integrity of the copy of the learning ability model;
releasing the learning ability model write lock;
and updating the learning ability model of the intelligent virtual digital animal according to the uploaded learning ability model copy.
7. The method of claim 4, wherein the learning ability model consists of a learning ability model structure definition, model parameters, and model parameter integrity check values; and obtaining a group of parameter sets, namely model parameters, through training and learning, and generating an integrity check value for each group of stable model parameters.
8. An animal robot for carrying out the method of claim 1, comprising a physical body and a computing system; the computing system downloads a learning ability model copy of the intelligent virtual digital animal from the cloud server, realizes personalized training of the learning ability model by performing scene interaction with a user, and uploads the trained learning ability model copy to the cloud server, so that the cloud server can update the learning ability model of the intelligent virtual digital animal according to the uploaded learning ability model copy; the animal robot is manufactured according to the structural body information of the intelligent virtual digital animal instance at the cloud server end, is used for physically mapping the intelligent virtual digital animal instance, and realizes the extension of the virtual digital world to the physical world; the physical body contains a physical representation of the properties and behavior of the intelligent virtual digital animal.
9. The animal robot of claim 8, wherein the physical body comprises a color, texture, programmable control manipulator, robotic arm, vision sensor; the computing system comprises an instance management agent module, a learning ability model copy module, an NPU module, a control logic integrated circuit and a storage module.
10. The animal robot of claim 9, wherein the instance management agent module communicates with the instance management module at the cloud server end through a network and is responsible for submitting an identity authentication request to the cloud hosting platform, downloading a learnable capacity model, verifying the downloaded learnable capacity model, installing the learnable capacity model as a local model copy of the animal robot physical object, scheduling a program for training and updating the learnable capacity model copy, and uploading the learnable capacity model copy after training and updating to the cloud; the NPU module is a special intelligent computing processor for accelerating the reasoning and training performance of the neural network, the control logic integrated circuit is a processor for realizing programmable control of the animal robot, and the storage module provides a rapid storage function for the copy of the learning ability model, training data and program operation.
11. A cloud server for implementing the method of claim 4, wherein an intelligent virtual digital animal program object instance cloud hosting platform is provided;
the intelligent virtual digital animal program object instance cloud hosting platform generates an intelligent virtual digital animal instance, and sends a learnable capacity model copy to an animal robot object according to a request of the animal robot object; the animal robot physical object is an animal robot manufactured according to an intelligent virtual digital animal example of a cloud server side;
the intelligent virtual digital animal program object instance cloud hosting platform receives the training-completed learning ability model copy uploaded by the animal robot object and updates the learning ability model of the intelligent virtual digital animal; the learning ability model copy after training is completed is a learning ability model copy after the user performs scene interaction with the animal robot real object and personalized training;
the cloud server comprises an instance management module, a virtual digital animal type module and an intelligent virtual animal instance module;
the instance management module manages the full life cycle of the intelligent virtual digital animal instance and the learning ability model management of the instance and is responsible for interactive communication with the instance management agent module running in the animal robot real object through a network;
the virtual digital animal type module comprises different animal type, and defines data structures of different animal types, including animal attributes, animal behaviors and frames of animal learning ability models;
the intelligent virtual animal instance module is a specific instantiation of intelligent virtual digital animal types and comprises a learning ability model, an attribute list and a behavior list.
12. A training learning system for personalizing a learning ability model of an intelligent virtual digital animal, comprising the animal robot of any one of claims 8-10, and the cloud server of claim 11.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911086750.4A CN110866588B (en) | 2019-11-08 | 2019-11-08 | Training learning method and system for realizing individuation of learning ability model of intelligent virtual digital animal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911086750.4A CN110866588B (en) | 2019-11-08 | 2019-11-08 | Training learning method and system for realizing individuation of learning ability model of intelligent virtual digital animal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110866588A CN110866588A (en) | 2020-03-06 |
| CN110866588B true CN110866588B (en) | 2023-05-02 |
Family
ID=69653228
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911086750.4A Active CN110866588B (en) | 2019-11-08 | 2019-11-08 | Training learning method and system for realizing individuation of learning ability model of intelligent virtual digital animal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110866588B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111461694B (en) * | 2020-03-17 | 2023-07-18 | 上海大学 | Dish identification and pricing system and method based on multi-level learning model |
| CN115699842A (en) * | 2020-07-31 | 2023-02-03 | Oppo广东移动通信有限公司 | Model management method, system, device, communication equipment and storage medium |
| CN112214748A (en) * | 2020-10-30 | 2021-01-12 | 支付宝(杭州)信息技术有限公司 | Identity recognition system, method and device |
| CN112668687B (en) * | 2020-12-01 | 2022-08-26 | 达闼机器人股份有限公司 | Cloud robot system, cloud server, robot control module and robot |
| CN112494956A (en) * | 2020-12-15 | 2021-03-16 | 四川长虹电器股份有限公司 | Simulation method and simulation system for converting articles into pets |
| CN113435126B (en) * | 2021-07-07 | 2024-02-02 | 魏天骐 | Knowledge sharing processing method, intelligent robot device, knowledge sharing system and task learning system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101174332A (en) * | 2007-10-29 | 2008-05-07 | 张建中 | Method, device and system for interactively combining real-time scene in real world with virtual reality scene |
| CA2941250A1 (en) * | 2014-03-19 | 2015-09-24 | Neurala, Inc. | Methods and apparatus for autonomous robotic control |
| CN108897848A (en) * | 2018-06-28 | 2018-11-27 | 北京百度网讯科技有限公司 | Robot interactive approach, device and equipment |
| WO2019057019A1 (en) * | 2017-09-20 | 2019-03-28 | 阿里巴巴集团控股有限公司 | Robot interaction method and device |
| CN110298212A (en) * | 2018-03-21 | 2019-10-01 | 腾讯科技（深圳）有限公司 | Model training method, Emotion identification method, expression display methods and relevant device |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090090305A1 (en) * | 2007-10-03 | 2009-04-09 | National University Of Singapore | System for humans and pets to interact remotely |
| JP6875513B2 (en) * | 2016-10-10 | 2021-05-26 | ディープマインド テクノロジーズ リミテッド | Neural network for selecting actions to be performed by robot agents |
| CN110139732B (en) * | 2016-11-10 | 2023-04-04 | 华纳兄弟娱乐公司 | Social robot with environmental control features |
| CN107139179B (en) * | 2017-05-26 | 2020-05-29 | 西安电子科技大学 | Intelligent service robot and working method |
| US11803764B2 (en) * | 2017-09-29 | 2023-10-31 | Sony Interactive Entertainment Inc. | Mobile and autonomous personal companion based on an artificial intelligence (AI) model for a user |
| CN109656529B (en) * | 2018-10-31 | 2021-05-25 | 北京大学 | Online customization method and system for deep learning of client |
| CN110378463B (en) * | 2019-07-15 | 2021-05-14 | 北京智能工场科技有限公司 | Artificial intelligence model standardization training platform and automatic system |
-
2019
- 2019-11-08 CN CN201911086750.4A patent/CN110866588B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101174332A (en) * | 2007-10-29 | 2008-05-07 | 张建中 | Method, device and system for interactively combining real-time scene in real world with virtual reality scene |
| CA2941250A1 (en) * | 2014-03-19 | 2015-09-24 | Neurala, Inc. | Methods and apparatus for autonomous robotic control |
| WO2019057019A1 (en) * | 2017-09-20 | 2019-03-28 | 阿里巴巴集团控股有限公司 | Robot interaction method and device |
| CN110298212A (en) * | 2018-03-21 | 2019-10-01 | 腾讯科技（深圳）有限公司 | Model training method, Emotion identification method, expression display methods and relevant device |
| CN108897848A (en) * | 2018-06-28 | 2018-11-27 | 北京百度网讯科技有限公司 | Robot interactive approach, device and equipment |
Non-Patent Citations (2)
| Title |
|---|
| Siau K等.Building trust in artificial intelligence, machine learning, and robotics.《Cutter business technology journal》.2018,第31卷(第2期),第47-53页. * |
| 王功利.基于HoloLens的虚拟交互宠物猫的设计与研究.《数字技术与应用》.2018,第36卷(第6期),第172-174页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110866588A (en) | 2020-03-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110866588B (en) | Training learning method and system for realizing individuation of learning ability model of intelligent virtual digital animal | |
| US11867599B2 (en) | Apparatus and methods for controlling attention of a robot | |
| US20240062081A1 (en) | Private artificial intelligence (ai) model of a user for use by an autonomous personal companion | |
| US8793205B1 (en) | Robotic learning and evolution apparatus | |
| CN108600379A (en) | A kind of isomery multiple agent Collaborative Decision Making Method based on depth deterministic policy gradient | |
| US20190065960A1 (en) | Continual selection of scenarios based on identified tags describing contextual environment of a user for execution by an artificial intelligence model of the user by an autonomous personal companion | |
| CN109145759A (en) | Vehicle attribute recognition methods, device, server and storage medium | |
| CN108028890A (en) | Crowdsourcing photography is managed in the wireless network | |
| CN104685516A (en) | Apparatus and methods for spiking neuron network learning | |
| CN111368789B (en) | Image recognition method, device, computer equipment and storage medium | |
| CN111768438B (en) | Image processing method, device, equipment and computer readable storage medium | |
| Shi et al. | A novel facial expression intelligent recognition method using improved convolutional neural network | |
| WO2024067884A1 (en) | Data processing method and related apparatus | |
| Zilly et al. | The ai driving olympics at neurips 2018 | |
| CN112990154B (en) | Data processing method, computer equipment and readable storage medium | |
| Ramachandruni et al. | Attentive task-net: Self supervised task-attention network for imitation learning using video demonstration | |
| US20190050881A1 (en) | Method and apparatus for rewarding reaction of simulation participant | |
| Song et al. | Types, locations, and scales from cluttered natural video and actions | |
| CN113159007A (en) | Gait emotion recognition method based on adaptive graph convolution | |
| Vogt | An overview of deep learning and its applications | |
| CN113282787B (en) | Personalized short video recommendation method and system based on reinforcement learning | |
| CN117635897B (en) | Three-dimensional object posture complement method, device, equipment, storage medium and product | |
| Wu et al. | Evolving neuromorphic systems on the ethereum smart contract platform | |
| Musse et al. | Interaction with virtual human crowds using artificial neural networks to recognize hands postures | |
| Chen | Robot learning from heterogeneous demonstration |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |