A robot cognitive development method based on ontology semantics

Abstract

The application discloses a robot cognitive development method based on ontology semantics, and comprises the following steps: constructing a robot article identification professional knowledge base based on attribute function and ontology information representation of article definition; information determination based on attribute discrimination and semantic search; and robot cognitive development based on attribute information addition. The application simulates the process of human memory, learning and cognition of articles based on an ontology semantic knowledge base, and through machine learning and sensor attribute information, the robot can actively cognize, learn, expand and accumulate learned knowledge and experience according to information data; the robot can continuously develop its cognitive ability through learning, automatically construct a robot article identification professional knowledge base of unknown articles, and based on a semantic structure of triplets, share knowledge and exchange operation logic between man and machine, so that the cognitive level of the robot is improved and the operation experience of an operator is improved.

Description

Technical Field

[0001] This invention belongs to the field of robot intelligent cognition, and specifically relates to a robot cognitive development method based on ontology semantics. Background Technology

[0002] In recent years, the application of artificial intelligence (AI) technology in the field of robotics has received increasing attention, aiming to enable robots to perform tasks more intelligently. When dealing with various tasks and dynamic environments in daily life, the main challenge for robots is how to quickly learn and memorize the characteristics of new objects, just like humans, and to effectively communicate and provide feedback to humans. Simultaneously, enabling robots to achieve cognitive recognition and learning development of objects, much like humans, is fundamental for them to perform complex tasks in unknown environments and is also an important technical method for improving the level of robot intelligence.

[0003] Currently, some studies have applied human-like cognitive development theories and brain mechanisms to the cognitive development of robots. For example, an autonomous cognitive development method based on incremental associative neural networks and dynamic audiovisual fusion, based on self-organizing neural networks, enables robots to autonomously develop object concepts and achieve audiovisual fusion; a self-organizing-reflective cognitive development method for robots with lifelong learning capabilities can autonomously identify and learn new types of modalities that suddenly appear online. However, these methods have the following drawbacks: (1) the process of humans endowing robots with relevant experiential knowledge is very complex, and the process logic and human-computer interaction are insufficient; (2) the knowledge content of the methods has low efficiency in utilizing prior knowledge, and insufficient scalability and sharing. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a robot cognitive development method based on ontology semantics, which can simulate the human cognitive memory and development process, realize the cognition and memory of new and old objects through the correlation between multiple attributes of objects, and realize the robot's cognitive recognition and learning development of objects.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] A robot cognitive development method based on ontology semantics includes the following steps:

[0007] (1) Based on the ontology information representation of attribute functions and item definitions, construct a professional knowledge base for robot item recognition: Under the guidance of cognitive experience, based on the relationship between the robot's attribute discrimination and classification ability and item recognition, construct an ontology triplet between attribute information and items;

[0008] (2) Information determination based on attribute discrimination and semantic search: For the input item information, attribute discrimination is performed respectively, and then the information fusion search based on the ontology is used to determine whether the item is an unknown item;

[0009] (3) Robot cognitive development based on attribute information addition: Store the attribute information of unknown items, and after supplementing the information of unknown items, expand it to the professional knowledge base of robot item recognition to realize the complete development and expansion of robot item cognition.

[0010] In step (1), the ontology information representation based on attribute functions and item definitions includes the following steps:

[0011] (1-1) Description and analysis of item attribute information based on cognitive experience: Based on people's cognitive information about items, items are described using various basic perceptual attributes, and information description and analysis of cognitive identification of items are carried out under the premise of the uniqueness of item attribute information; the uniqueness of item attribute information means that there are no identical elements in the large set of item descriptions, satisfying the mutual dissimilarity of sets;

[0012] (1-2) Establishment of a professional knowledge base for robot item recognition based on knowledge graph technology: On the Protege visualization platform, based on the skeleton construction method, with the robot as the ontological object of cognition and perception, and each attribute information and item as the first-level subclass, based on the description analysis in step (1-1), add item instance objects and attribute information instances based on the description instance objects to establish a professional knowledge base for robot item recognition.

[0013] In steps (1-2), a set R is established with the robot as the cognitive subject. Set R contains items W, attributes S, and a large set P describing the relationships between items and attributes. Item instances (A, B, C, ...) are added, such that W = {item A, item B, item C, ...}. Simultaneously, based on the attribute category, attributes are divided into attribute 1, attribute 2, attribute 3, ..., such that S = {attribute 1, attribute 2, attribute 3, ...}, and related attribute instances a, b, c, ... are added, such that S = {a, b, c, ...}. Furthermore, based on the actual correspondence between attribute instances and attributes, such as a belonging to attribute 1, an element dependency relationship is constructed: <attribute 1, hasindividual, a>. The remaining attribute instances are similarly assigned dependency relationships. Then, the root... Based on the item description relationships obtained in step (1-1), ontology triples are used for description. An ontology triple refers to an item, an attribute, and the relationship between items and attributes. An ontology triple <item A, attribute 1, a> indicates that the state of attribute 1 of item A is a, meaning a triple describing item A is A1. Based on i different attributes, i sets of triples describing item A can be obtained, i.e., the description set of item A is A = {A1, A2, A3, ..., Ai}. Similarly, attribute description sets for items B, C, ... can be constructed, forming a large set of item descriptions P = {A, B, C, ...}. Sets W and S are connected using predicate descriptive terms, i.e., based on the large set of item descriptions P, sets W and S are mapped to obtain the robot cognitive subject set R = {W, S, P}. Set R is the set representation of the robot's item recognition professional knowledge base.

[0014] In step (2), the information determination based on attribute discrimination and semantic search includes the following steps:

[0015] (2-1) Attribute information acquisition based on robot object recognition professional knowledge base: using machine learning, statistical learning, sensor information source acquisition and other methods, the robot can acquire human-like cognitive attributes such as color attributes, shape attributes, three-dimensional reconstruction information, weight attributes, and texture attributes;

[0016] (2-2) Semantic search rule formulation based on information input: When item information is input, after the attribute capability unit judges it, its attribute information signals are all converted into triplet data records; at the same time, SPARQL query is used to merge the attribute information using "AND" relationship, and then the search is performed in the established robot item recognition professional knowledge base based on the search language of ontology knowledge base to obtain the number of items N that meet the conditions and the search results F = {f1,...,fN}. If the number of items N = 1, it can be determined that F is the unique corresponding result object of the unknown item X, and the output is: "Item is F", indicating that the robot has learned the item; if N > 1, it means that there are multiple items in the cognitive base that meet the attribute description, and the attribute needs to be added for judgment, and the output is "Item may be: F, attribute judgment needs to be added"; if N < 1, it means that there is no item that completely meets the description in the existing robot item recognition professional knowledge base, which is an unknown item, and the robot needs to supplement the learning.

[0017] In step 2, the information determination and description recording of unknown items can be described as follows: Obtain the attribute information of unknown item X, and form the corresponding ontology triple <item X, attribute 1, x>, denoted as X1. Similarly, obtain the description set X = {X1, X2, X3, ..., Xi} of item X. Then, based on the ontology triple search rules of SPARQL, perform strict matching between item description sets. If the attribute description set of item F can be found in the large set of item descriptions P that is completely equal to the description set of item X, i.e., F = X, where F1 = X1, F2 = X2, F3 = X3, ..., Fi = Xi, then item X is item D. Based on the input of prior knowledge, the robot can know the name and related information of item X. When no item description set can be found in the large set of item descriptions P that is strictly equal to the set of X, then item X can be determined to be an unknown item of the robot, and needs to be added to the robot's item recognition professional knowledge base.

[0018] In step (3), the information-added-based robot cognitive development includes the following steps:

[0019] (3-1) Automatic recording and addition of unknown items: A new temporary database is added to record the attribute information of unknown items and assign numbers to the unknown items; at the same time, based on the automatic construction technology of knowledge graph, the number is used as the name of the unknown item and the attribute information is used as the instance information. Each instance is added to the robot's professional knowledge base for item recognition to complete short-term cognitive development.

[0020] (3-2) Information supplementation and cognitive development achievement: After the short-term task is completed, the operator can identify and modify the name of the unknown item based on the attribute information of the unknown item, form long-term memory, and achieve the goal of complete cognitive development.

[0021] Compared with existing technologies, this invention has the following advantages and effects: Based on an ontology semantic knowledge base, this invention mimics the human process of remembering, learning, and recognizing objects. Through machine learning and sensor attribute information, it enables the robot to actively recognize, learn, expand, and accumulate learned knowledge and experience based on information data. This allows the robot to continuously develop its cognitive abilities through learning, automatically construct a professional knowledge base for robot object recognition of unknown objects, and, based on the semantic structure of triples, facilitate knowledge sharing and operational logic communication between humans and machines, improving the robot's cognitive level and enhancing the operator's experience. Simultaneously, based on the semantic records of triples during the process, humans can understand the evaluation criteria and specific procedures of the robot's cognitive decision-making process. Feedback allows for accurate and timely identification of specific problems encountered by the robot, enabling problem-solving and further enhancing the robot's intelligence level. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the robot cognitive development method based on ontology semantics according to the present invention.

[0023] Figure 2 This is a schematic diagram of the original knowledge base in Example 1.

[0024] Figure 3 This is the flowchart for determining the new fruit in Example 1.

[0025] Figure 4 This is a schematic diagram of the expanded knowledge base in Example 1.

[0026] Figure 5 This is a schematic diagram of the knowledge base after development is achieved in Example 1. Detailed Implementation

[0027] To facilitate understanding of the present invention, specific embodiments will be described in detail below. These embodiments will help those skilled in the art to further understand the present invention; however, they are not intended to limit the invention in any way. It should be noted that those skilled in the art can make various modifications and improvements to the present invention without departing from its conceptual framework, and these modifications and improvements all fall within the scope of protection of the present invention.

[0028] Example 1

[0029] This invention relates to a robot cognitive development method based on ontology semantics, such as... Figure 1 As shown, the robot's internal information processing flow includes the following steps:

[0030] Step 1: Acquisition and description of the attribute information of the items to be identified: Based on the professional knowledge base for robot item recognition constructed by attribute functions and item definitions, an item attribute information acquisition platform is built. At the same time, the item triplet description is performed based on the acquisition information.

[0031] Step 2: Based on the triplet description of the item, perform an item matching search within the robot's professional knowledge base for robot item recognition to determine whether the item is an unknown item.

[0032] Step 3: If the item to be identified is not an unknown item, that is, the item information can be found in the robot's item recognition professional knowledge base, then the item information can be output and the item recognition process ends; if the item to be identified is an unknown item, then the attribute information is stored in a temporary database with a number, and then proceed to Step 4.

[0033] Step 4: Add the attribute information of the unknown item to the robot's object recognition professional knowledge base, replace the unknown item with its identification number, complete the temporary cognitive development, and proceed to Step 5. That is, if the unknown item's identification number in the temporary database is 5, then when the knowledge base is expanded, the robot will recognize the item as the previously encountered item number 5 when it encounters it again.

[0034] Step 5: When the robot interacts with the operator during the operation process, the operator can identify unknown items based on the previous attribute descriptions, provide the robot with the correct item name information, and complete the entire cognitive development process.

[0035] This invention simulates the cognitive development process of humans. It performs memory matching search on the items to be distinguished. If the item is unknown, it remembers the relevant attributes and characteristics and performs descriptive memory, which constitutes temporary cognitive development. After the information is improved and supplemented, the complete cognitive development process of recognizing the new item is completed.

[0036] Furthermore, this embodiment takes the cognitive development process of a robot towards an unknown fruit as an example, including the following steps:

[0037] Step 1: The information that distinguishes fruits lies in their color, shape, volume, three-dimensional information, density, texture, etc. We need to build a professional knowledge base for the representation analysis of fruits based on the above attribute information and for robot object recognition.

[0038] Since the instance environment only contains ripe apples, ripe bananas, and one unknown fruit, for the sake of brevity, we only choose color and shape as the attribute description information of the fruit. The color of a ripe apple is red and the shape is round; the color of a ripe banana is yellow and the shape is long and narrow.

[0039] Subsequently, a fruit attribute information knowledge base was built based on the Protege platform, with the robot as the cognitive and perceptual ontology object. Fruit, shape, and color were used as first-level subclasses. Instances were added under the fruit subclass: ripe apple, ripe banana; under the color subclass: yellow, red; and under the shape subclass: oblong, round. The predicate phrases `has_color` and `has_shape` were added to construct ontology triples between attribute information and item information. Specifically, `<ripe apple, has_color, red>`, `<ripe apple, has_shape, round>`, `<ripe banana, has_color, yellow>`, and `<ripe banana, has_shape, oblong>`.

[0040] The robot's object recognition knowledge base includes an object set W = {ripe apple, ripe banana}, an attribute set S = {shape, color} = {striped, round, yellow, red}, a description set A = {<ripe apple, has_color, red>, <ripe apple, has_shape, round>}, a description set B = {<ripe banana, has_color, yellow>, <ripe banana, has_shape, striped>}, and a large set of object descriptions P = {A, B} = {<ripe apple, has_color, red>, <ripe banana, has_shape, round>}, <ripe apple, has_color, red>, <ripe apple, has_shape, round>}, and a large set of object descriptions P = {A, B} = {<ripe apple, has_color, red>, <ripe apple, has_shape, round>}, < ... The robot's cognitive subject set R = {W, S, R} = {ripe apple, ripe banana, shape, color, elongated, round, yellow, red, <ripe apple, has_color, red>, <ripe apple, has_shape, round>, <ripe banana, has_color, yellow>, <ripe banana, has_shape, elongated>} is constructed as follows: Figure 2 As shown.

[0041] Step 2: Based on the content of Step 1, we know that the robot needs to have two abilities in the environment: color attribute and shape attribute. After enabling the robot to have these two attribute abilities, we judge the input information (in this example, since only the shape and color attribute information is described and constructed, the input is image information). We use an "AND" relationship to search the knowledge base for shape information and color information to see if they meet the known fruit description. If they do, we get the fruit name; if not, we output the feedback to the human.

[0042] For obtaining the robot's color attributes, we used the SVM method to train on standard shape images (circles, rectangles, rhombuses, pentagons, etc.). After the image was input, the SVM distinguished and judged the shape of the object in the image, and we could determine the shape name. For obtaining the robot's color attributes, we used mathematical statistics to statistically analyze the color of the object after the image was corrected by a standard color chart. The color was represented by the range of RGB values ​​obtained from the image. Finally, the range of red was determined to be B∈[0, 64], G∈[0, 47], R∈[20, 152], and the range of yellow was determined to be B∈[0, 114], G∈[92, 200], R∈[100, 208].

[0043] After obtaining the item attribute information, we use the SPARQL query language to search within the knowledge base. Assume the computer obtains an image of fruit X with color C and shape S. That is, inputting color C and shape S into the knowledge base, the search outputs fruit X. First, searching using color C and shape S as common conditions yields the number of fruits N that satisfy the conditions and the search results F = {f1,...,fN}. If the number of fruits N = 1, we can determine that F is fruit X and output: "Fruit is F". If N > 1, it means that different fruits with the same color and shape exist in the knowledge base, requiring further attribute judgment, and the output is "Fruit may be: F, attribute judgment needs to be added". If N < 1, it means that there is no fruit in the knowledge base that completely meets the conditions, or that an attribute judgment error has occurred, resulting in no result being obtained, and the output is "New fruit found, fruit color is C, shape is S". The process is as follows: Figure 3 As shown.

[0044] Step 3: When a new fruit is identified, the robot stores and numbers its attribute information. For example, if an unknown fruit image is input: round shape, yellow color, the image information is saved to the "New Items" folder, and the new item is numbered: 1, round shape, yellow color. Then, the number, name, and related attribute information are automatically added to the ontology knowledge base. An instance "New Fruit 1" is added under the "Fruit" subclass, and the attribute information is connected by a predicate verb to form a triple: <New Fruit 1, has_color, yellow>, <New Fruit 1, has_shape, round>. When the robot encounters this new fruit again, it can recognize it as New Fruit 1. The expanded knowledge base includes... Figure 4 As shown.

[0045] After receiving prompts from the robot, the name of the new fruit 1 can be changed. Viewing the image of the new fruit 1 reveals that it is yellow and round, representing a pear. Therefore, the name of the new fruit 1 is changed to "pear" in the knowledge base. This completes the developmental recognition of the new fruit 1, realizing the developmental expansion of the robot's humanoid object recognition. The final knowledge base is as follows... Figure 5 As shown.

[0046] The above description is merely an embodiment of the present invention, but the implementation of the present invention is not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention are equivalent substitutions and are included within the protection scope of the present invention.

Claims

1. A robot cognitive development method based on ontology semantics, characterized in that... Includes the following steps: (1) Based on the ontology information representation of attribute functions and item definitions, construct a professional knowledge base for robot item recognition, including the following steps: (1-1) Description and analysis of item attribute information based on cognitive experience: Based on people's cognitive information about items, the items are described using various basic perceptual attributes, and the information description and analysis of cognitive identification of items are carried out under the premise of the uniqueness of item attribute information; the uniqueness of item attribute information means that there are no identical elements in the large set of item descriptions, satisfying the mutual dissimilarity of sets; (1-2) Establishment of a professional knowledge base for robot item recognition based on knowledge graph technology: A set R is established with the robot as the cognitive subject. Set R contains items W, attributes S, and a large set P of item descriptions of the relationship between items and attributes. Item instances, i.e., items A, B, C, ..., are added, such that W = {item A, item B, item C, ...}. Simultaneously, based on the category of attributes, attributes are divided into attribute 1, attribute 2, attribute 3, ..., such that S = {attribute 1, attribute 2, attribute 3, ...}, and related attribute instances a, b, c, ... are added, such that S = {a, b, c, ...}. Furthermore, based on the actual correspondence between attribute instances and attributes, i.e., a belongs to attribute 1, an element dependency relationship is constructed: <attribute 1, has Individual, a>; the other attribute instances are also assigned subordinate relationships in the same way; then, based on the item description relationship obtained in step (1-1), the ontology triplet description is performed. The ontology triplet refers to the expression of the relationship between item, attribute, and item and attribute. The ontology triplet <item A, attribute 1, a> indicates that the attribute 1 state of item A is a, that is, a triplet describing item A is A1. Based on i different attributes, i sets of triplets describing item A are obtained, that is, the description set of item A is A={A1, A2, A3, ..., Ai}; similarly, the attribute description sets of item B, item C, ... are constructed to form a large set of item descriptions P={A, B, C, ...}; the set W and the set S are connected by predicate descriptive words, that is, the set W and the set S are mapped based on the large set of item descriptions P to obtain the robot cognitive subject set R={W, S, P}. The set R is the set representation of the robot item recognition professional knowledge base; (2) Information determination based on attribute discrimination and semantic search: For the input item information, attribute discrimination is performed respectively, and then the information fusion search based on the ontology is used to determine whether the item is an unknown item; (3) Robot cognitive development based on attribute information addition: Store the attribute information of unknown items, and after supplementing the information of unknown items, expand it to the professional knowledge base of robot item recognition to realize the complete development and expansion of robot item cognition.

2. The ontology-based robot cognitive development method according to claim 1, characterized in that: In step (2), the information determination based on attribute discrimination and semantic search includes the following steps: (2-1) Attribute information acquisition based on robot object recognition professional knowledge base: using machine learning, statistical learning and sensor information source acquisition methods to enable robots to acquire human-like cognitive attribute capabilities; (2-2) Semantic search rule formulation based on information input: When item information is input, after the attribute capability unit judges it, the attribute information signals are all converted into triplet data records; at the same time, SPARQL query is used to merge the attribute information using "AND" relationship, and then the search is performed in the established robot item recognition professional knowledge base based on the search language of the ontology knowledge base to obtain the number of items N that meet the conditions and the search results F={f1,...,fN}. If the number of items N=1, that is, F is judged to be the unique corresponding result object of the unknown item X, and the output is: "Item is F", indicating that the robot has learned the item; if N>1, it means that there are multiple items that meet the attribute description in the robot item recognition professional knowledge base, and the attribute needs to be added for judgment, and the output is "Item may be: F, attribute judgment needs to be added"; if N<1, it means that there are no items that completely meet the description in the existing robot item recognition professional knowledge base, which are unknown items, and the robot needs to supplement the learning.

3. The robot cognitive development method based on ontology semantics according to claim 1, characterized in that: In step (2), the information determination and description of unknown items are described as follows: obtain the attribute information of unknown item X, and form the corresponding ontology triple <item X, attribute 1, x>, denoted as X1. Similarly, obtain the description set X = {X1, X2, X3, ..., Xi} of item X. Then, based on the ontology triple search rules of SPARQL, perform strict matching between item description sets. If the attribute description set of item F is found to be completely equal to the description set of item X in the large set of item descriptions P, i.e., F = X, where F1 = X1, F2 = X2, F3 = X3, ..., Fi = Xi, then it means that item X is item F. Based on the input of prior knowledge, the robot knows the name and related information of item X. When no item description set can be found in the large set of item descriptions P that is strictly equal to the set of X, then item X is determined to be an unknown item of the robot and needs to be added to the robot's item recognition professional knowledge base.

4. The robot cognitive development method based on ontology semantics according to claim 1, characterized in that: In step (3), the robot cognitive development based on attribute information includes the following steps: (3-1) Automatic recording and addition of unknown items: A new temporary database is added to record the attribute information of unknown items and assign numbers to the unknown items; at the same time, based on the automatic construction technology of knowledge graph, the number is used as the name of the unknown item and the attribute information is used as the instance information. Each instance is added to the robot's professional knowledge base for item recognition to complete short-term cognitive development. (3-2) Information supplementation and cognitive development achievement: After the short-term task is completed, the operator identifies and modifies the name of the unknown item based on the attribute information of the unknown item, forming long-term memory and achieving the goal of complete cognitive development.

Images