An improved intelligent cognitive robot cognitive module based on ACT-R

Abstract

The application relates to the field of intelligent cognitive technology and discloses an improved intelligent cognitive robot cognitive module based on ACT-R, which comprises cognitive modeling intelligent unit modules, feature processing intelligent unit modules, understanding calculation intelligent unit modules, reasoning prediction intelligent unit modules and association recommendation intelligent unit modules which are connected in sequence. The application solves the problems that the existing cognitive calculation does not involve human emotions and implicit knowledge and cannot more truly simulate the human cognitive decision-making process.

Description

TECHNICAL FIELD

[0001] The application relates to the field of intelligent cognitive technology, and particularly relates to an improved intelligent cognitive robot cognitive module based on ACT-R. BACKGROUND

[0002] Since the deep blue computer of IBM defeated the then world chess champion Kasparov in 1996, cognitive intelligence has attracted people's attention. Cognitive intelligence is the basic science of intelligent cognitive robots and is a high-level stage of the development of intelligent science. It is based on the human cognitive system and mode and aims to imitate the core wisdom and ability of human beings. With the development of artificial intelligence and the increasingly serious global population aging problem, the application prospect of intelligent cognitive robots is particularly wide. Therefore, the research on the core "intelligent cognitive module" of intelligent cognitive robots is particularly important and urgent. The learning object of the intelligent cognitive robot is human being, and the analysis, decision, planning, deduction, route planning and action planning of the intelligent cognitive robot cannot be separated from cognitive computing, and the foregoing work cannot be separated from a reasonable and effective intelligent cognitive framework.

[0003] ACT-R (Adaptive Control of Thought-Rational) is a cognitive architecture established by American artificial intelligence experts and psychologists such as Aderson, which is a theory used to simulate and try to understand human cognition, including perceptual-motor modules, goal modules, declarative memory modules and other modules. ACT-R tries to understand how humans organize knowledge and produce intelligent behavior, and its goal is to enable systems to perform various cognitive tasks of humans, such as capturing human perception, thought and behavior. The core of ACT-R is the descriptive knowledge module and the central production system. The descriptive knowledge module stores the individual's accumulated long-term and unchanging knowledge, including basic facts, professional knowledge, etc. The central production system stores the individual's procedural knowledge, which is in the form of conditional-action (production) rules. When certain conditions are met, the corresponding action will be executed by the corresponding module, and the continuous triggering of production rules can ensure the cooperation of each module to simulate the continuous cognitive process made by the individual. ACT-R has undergone several rounds of important version development. In version 4.0, it successfully modeled the cognitive phenomena of the first two domains of the unified domain (i.e., problem solving, decision making, routine action, memory, learning and skill). In version 5.0, it modeled perception and motor behavior. ACT-R has now entered version 6.0 and supports different system operation platforms. As an ACT-R theory with a large amount of experimental information, it has been widely applied in the field of psychology research and can be directly used in the research work of intelligent cognitive robots. However, in the process of human cognition, not only declarative knowledge and procedural knowledge are involved, but also human emotional factors and implicit knowledge. The current cognitive computing in the field of artificial intelligence does not involve human emotions and implicit knowledge, and the Watson of IBM's cognitive solution spokesperson also does not involve these two aspects. SUMMARY

[0004] In order to overcome the shortcomings of the prior art, the present application provides an improved intelligent cognitive robot cognitive module based on ACT-R, which solves the problems of the prior art that cognitive computing does not involve human emotions and implicit knowledge, and it is difficult to more realistically simulate human cognitive decision-making processes.

[0005] The technical scheme adopted by the present application to solve the above problems is:

[0006] An improved intelligent cognitive robot cognitive module based on ACT-R, comprising a cognitive modeling intelligent unit module, a feature processing intelligent unit module, an understanding calculation intelligent unit module, a reasoning prediction intelligent unit module, and an association recommendation intelligent unit module connected in sequence; wherein the cognitive modeling intelligent unit module is used to: build a purpose model through the cognitive modeling intelligent unit module, the purpose model at least including an information processing model, the information processing model indicating a computer intelligent processing algorithm model selected respectively for different information; the feature processing intelligent unit module is used to: perform feature extraction of information layers on the acquired multi-dimensional information and feature extraction of signal layers on the acquired multi-modal interactive information by using the selected information processing model; the understanding calculation intelligent unit module is used to: perform understanding calculation according to multi-dimensional signal features and multi-modal information features; the reasoning prediction intelligent unit module is used to: perform reasoning prediction according to the results of understanding calculation to obtain intention information; and the association recommendation intelligent unit module is used to: make action planning in advance for the intelligent cognitive robot according to the intention information.

[0007] As a preferred technical solution, the purpose model further includes a cognitive model, the cognitive model indicating a cognitive understanding model selected by the intelligent cognitive robot according to different interactive objects and different interactive tasks.

[0008] As a preferred technical solution, the purpose model further includes an emotion model, the emotion model indicating an emotion information analysis model selected by the intelligent cognitive robot according to an interactive task for an interactive object.

[0009] As a preferred technical solution, the purpose model further includes a thinking model, the thinking model indicating an implicit knowledge representation model for different tasks selected by the intelligent cognitive robot according to different interactive objects and different interactive tasks.

[0010] As a preferred technical solution, the understanding calculation intelligent unit module can be used for emotion calculation, environment state calculation, and cognitive target calculation, wherein the emotion calculation indicates calculation of an emotional state of an interactive object according to acquired emotion information, the environment state calculation indicates comprehensive calculation of an environment situation related to a current interactive task according to acquired various environment information, and the cognitive target calculation indicates calculation of emotion, environment, and perception comprehensive condition information related to a target to be achieved according to a target to be achieved by an interactive task.

[0011] As a preferred technical scheme, the memory solidification intelligent unit module is further connected with the cognitive modeling intelligent unit module and the understanding calculation intelligent unit module, and is used for converting expression information obtained by the use model into a vector.

[0012] As a preferred technical scheme, the decision judgment intelligent unit module is further connected with the reasoning prediction intelligent unit module, and is used for realizing high-order cognitive ability of the intelligent cognitive robot according to intention information obtained by the reasoning prediction intelligent unit module.

[0013] As a preferred technical scheme, the feedback learning intelligent unit module is further connected with the cognitive modeling intelligent unit module, the understanding calculation intelligent unit module and the decision judgment intelligent unit module, and is used for learning rules and modes according to feedback information of the cognitive modeling intelligent unit module and the understanding calculation intelligent unit module, and constantly optimizing high-order cognitive ability.

[0014] As a preferred technical scheme, the use model further comprises a comprehensive perception model and an environment model, the comprehensive perception model refers to a perception model selected by the intelligent cognitive robot according to characteristics of perception conditions of an interactive object, and the environment model refers to an environment model provided by the intelligent cognitive robot according to characteristics of an environment in which the interactive object is located and an environment required by an interactive task; the feedback learning intelligent unit module can learn rules and modes according to feedback information of the comprehensive perception model and the environment model, and constantly optimize high-order cognitive ability.

[0015] As a preferred technical scheme, the comprehensive perception model refers to a perception model selected by the intelligent cognitive robot according to characteristics of perception conditions of an interactive object, and the perception model comprises a dynamic combination of touch perception, hearing perception, electroencephalogram perception, physiological sign perception, force feedback perception and environment perception.

[0016] Compared with the prior art, the application has the following beneficial effects:

[0017] (1) The intelligent robot cognitive decision of the application is more real and accurate, and more factors influencing final behavior decision in human cognitive process, such as human emotion, implicit knowledge, explicit knowledge and surrounding environment, are considered, so that the intelligent robot is closer to real human judgment in path planning, intention analysis and cognitive decision.

[0018] (2) The memory solidification intelligent unit module can not only solidify and store declarative knowledge, but also solidify and store implicit knowledge, thereby supporting the improvement of intelligent cognitive level; the memory solidification intelligent unit module converts the expression information obtained through the aforementioned cognitive model, emotion model, thinking model and each purpose model into vectors, that is, converts signal information and text information into vectors for comprehensive operation, and records and stores the vectors, so that the solidification storage of implicit knowledge and the solidification storage of declarative knowledge can be realized;

[0019] (3) The intelligent cognitive robot reasoning prediction result accuracy can be improved by the application, the human emotion information and the surrounding environment information are considered at the same time, so that the reasoning prediction result is closer to the actual demand;

[0020] (4) The intelligent cognitive robot cognitive learning ability is improved, the feedback learning intelligent unit module learns and summarizes certain rules and modes from the feedback information of the outside world, and the decision-making intelligent unit module is continuously optimized, so that the cognitive growth of the intelligent cognitive robot is realized. BRIEF DESCRIPTION OF DRAWINGS

[0021] Fig. 1 is a composition schematic view of an improved intelligent cognitive robot cognitive module based on ACT-R according to the application;

[0022] Fig. 2 is a working flow chart of an improved intelligent cognitive robot cognitive module based on ACT-R according to the application. DETAILED DESCRIPTION

[0023] The application will be further described in detail below in combination with embodiments and drawings, but the embodiments of the application are not limited thereto.

[0024] Embodiment 1

[0025] As shown in Figs. 1-2 , the application proposes and constructs an improved intelligent cognitive robot cognitive module based on ACT-R, and simulates the human thinking process in the intelligent cognitive module, and adds intelligent unit modules such as cognitive modeling, feature processing, reasoning prediction, association recommendation and feedback learning.

[0026] The application aims to solve the problems in the prior art, provide a robot cognitive module capable of simulating human cognitive decision-making process and comprehensively considering human emotional factors and implicit knowledge, and support the intelligent cognitive robot to provide behaviors and judgments more in line with real scene requirements in the planning and decision-making process.

[0027] The above-mentioned object of the present application can be achieved by the following measures, an improved intelligent cognitive robot cognitive module based on ACT-R, comprising: cognitive modeling, memory solidification, feature processing, understanding calculation, reasoning prediction, association recommendation, decision judgment and feedback learning intelligent unit module parts.

[0028] The intelligent cognitive module (i.e. an improved intelligent cognitive robot cognitive module based on ACT-R according to the present application) can first construct cognitive models, emotion models, thinking models (oriented to implicit knowledge), comprehensive perception models, environment models, declarative knowledge models and information processing models through the cognitive modeling intelligent unit module, and the related models provide input standards for the subsequent feature processing, understanding calculation, memory solidification and other intelligent unit module processing links; wherein the cognitive model refers to the cognitive understanding model selected by the intelligent cognitive robot according to different interactive objects and different interactive tasks; the emotion model refers to the emotion information analysis model selected by the intelligent cognitive robot for the interactive object according to the interactive task; the thinking model refers to the implicit knowledge representation model for different tasks selected by the intelligent cognitive robot according to different interactive objects and different interactive tasks, which is mainly composed of a large number of task-related data trained models; the comprehensive perception model refers to the perception model selected by the intelligent cognitive robot according to the perception condition characteristics of the interactive object, which contains the dynamic combination of touch, hearing, brain electrical perception, physiological sign perception, force feedback perception and environmental perception; the environment model refers to the environment model provided by the intelligent cognitive robot according to the environment of the interactive object and the environment characteristics required by the interactive task, which is dynamically combined by season, indoor and outdoor temperature, humidity and light intensity; the declarative knowledge model refers to the representation model of knowledge such as processing flow, common sense and formula related to the interactive task which can be described; the information processing model refers to the computer intelligent processing algorithm model selected for different perception information, cognitive information and knowledge information;

[0029] The feature processing intelligent unit module uses the information processing model selected by the intelligent cognitive robot to perform feature extraction on the physiological information, environmental information and brain signal information obtained from the intelligent cognitive robot perception module, and the multi-modal interactive information obtained from the human-computer interaction module in the information layer and the signal layer;

[0030] In this process, the memory solidification intelligent unit module not only stores the structured knowledge, but also stores the implicit knowledge; the memory solidification intelligent unit module converts the expression information obtained through the aforementioned cognitive model, emotion model and thinking model into vectors, i.e. converts the signal information and text information into vectors for comprehensive operation, so as to realize the solidification storage of implicit knowledge and the solidification storage of declarative knowledge;

[0031] The understanding computing intelligence unit module is mainly used for emotion computing, environment state computing and cognitive target computing, wherein the emotion computing refers to computing the emotion state of an interactive object according to acquired emotion information, the environment state computing refers to comprehensively computing the environment situation related to the current interactive task according to acquired various environment information, and the cognitive target computing refers to computing the comprehensive condition information such as emotion, environment and perception related to the target to be achieved according to the target to be achieved by the interactive task; the computed information is used for reasoning and prediction intelligence unit modules of relationship, trend, state and intention;

[0032] According to the intention information obtained by the reasoning and prediction intelligence unit module, the association recommendation intelligence unit module can make action planning for the intelligent cognitive robot in advance;

[0033] Meanwhile, the decision and judgment intelligence unit module can realize the high-order cognitive ability of behavior control and decision and judgment of the robot;

[0034] In order to improve the cognitive learning ability of the intelligent cognitive robot, the feedback learning intelligence unit module learns and summarizes certain rules and modes from the feedback information from the outside world, and continuously optimizes the decision and judgment intelligence unit module, so as to realize the cognitive growth of the intelligent cognitive robot.

[0035] Compared with the prior art, the present application has the following beneficial effects:

[0036] The cognitive decision of the intelligent robot is more real and accurate. More factors affecting the final behavior decision in the human cognitive process such as human emotion, implicit knowledge, explicit and declarative knowledge, and surrounding environment are comprehensively considered, so that the intelligent robot is closer to the real judgment of human in path planning, intention analysis and cognitive decision;

[0037] The memory solidification intelligence unit module can not only solidify and store the declarative knowledge, but also solidify and store the implicit knowledge, thereby supporting the improvement of the intelligent cognitive level. The memory solidification intelligence unit module converts the expression information obtained by the foregoing cognitive model, emotion model and thinking model into vectors, i.e. converts the signal information and text information into vectors for comprehensive operation, so as to record and store the vectors, which can realize the solidification storage of the implicit knowledge and the solidification storage of the declarative knowledge;

[0038] The present application can enhance the accuracy of the reasoning and prediction result of the intelligent cognitive robot. The present application simultaneously considers human emotion information and surrounding environment information, so that the reasoning and prediction result is closer to the actual demand;

[0039] The application improves the cognitive learning ability of the intelligent cognitive robot. The feedback learning intelligent unit module learns certain rules and patterns from the feedback information from the outside world, and continuously optimizes the decision-making judgment intelligent unit module, thereby realizing the cognitive growth of the intelligent cognitive robot.

[0040] Embodiment 2

[0041] As shown in Figs. 1-2 As a further optimization of embodiment 1, on the basis of embodiment 1, the present embodiment further comprises the following technical features:

[0042] Referring to Fig. 1 In the preferred embodiments described below, an improved intelligent cognitive robot cognitive module based on ACT-R comprises: cognitive modeling, memory consolidation, feature processing, understanding calculation, reasoning prediction, association recommendation, decision-making judgment and feedback learning intelligent unit module parts.

[0043] Referring to Fig. 2 The main purpose of the cognitive modeling intelligent unit module is to first provide the construction ability of cognitive models, emotional models, thinking models (oriented to implicit knowledge), comprehensive perception models, environmental models, declarative knowledge models and information processing models through the cognitive modeling intelligent unit module, and the relevant models provide input standards for subsequent feature processing, understanding calculation, memory consolidation and other intelligent unit module processing links. For example, taking man-machine confrontation as an example, in the face of interactive objects in combat, the cognitive modeling intelligent unit module automatically selects the action guidance model in the confrontation process with the intelligent cognitive robot, and other model contents are shown in the following table:

[0044]

[0045] The feature processing intelligent unit module is used to extract features from the physiological information, environmental information, brain signal information and multi-modal interaction information obtained from the combat opponent in the information layer and signal layer;

[0046] In this process, the knowledge that can be structured and described will be stored through the memory consolidation intelligent unit module, so that the information of the opponent country, military, gender, age, height, weight, etc. The memory consolidation intelligent unit module will convert the expression information about the combat opponent obtained through the aforementioned cognitive model, emotional model, thinking model and other purpose models into vectors, i.e. convert the signal information and text information into vectors for comprehensive operation, record and store in vectors, and realize the solidification storage of implicit knowledge.

[0047] The understanding calculation intelligent unit module calculates the emotions of the person in confrontation with the intelligent cognitive robot, the state of the confrontation environment, and the goal of this confrontation action in the combat process; the calculated information is used for reasoning prediction intelligent unit module in relation, trend, state and intention, etc.

[0048] According to the combat confrontation intention information derived by the reasoning prediction intelligent unit module, the association recommendation intelligent unit module can plan in advance for the subsequent action of the intelligent cognitive robot;

[0049] Meanwhile, the decision-making judgment intelligent unit module can realize the high-order cognitive ability of the combat behavior control and combat decision-making judgment of the intelligent robot;

[0050] In order to improve the man-machine confrontation cognitive learning ability of the intelligent cognitive robot, the feedback learning intelligent unit module learns and summarizes certain rules and patterns from the feedback information from the outside world, and continuously optimizes the decision-making judgment intelligent unit module, so as to realize the cognitive growth of the intelligent cognitive robot.

[0051] As described above, the present application can be better implemented.

[0052] All features disclosed in the embodiments of the present specification, or all steps in the methods or processes impliedly disclosed, can be combined and / or extended, replaced, unless mutually exclusive features and / or steps are mutually exclusive.

[0053] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. According to the technical essence of the present application, any simple modification, equivalent replacement and improvement of the above embodiment within the spirit and principles of the present application are still within the protection scope of the technical scheme of the present application.

Claims

1. An improved intelligent cognitive robot cognitive module based on ACT-R, characterized in that, The cognitive modeling intelligent unit module, the feature processing intelligent unit module, the understanding calculation intelligent unit module, the reasoning prediction intelligent unit module and the association recommendation intelligent unit module are sequentially connected. The use model further includes a cognitive model, the cognitive model refers to a cognitive understanding model selected by the intelligent cognitive robot according to different interactive objects and different interactive tasks. The use model further includes an emotion model, the emotion model refers to an emotion information analysis model selected by the intelligent cognitive robot according to an interactive task for an interactive object. The use model further includes a thinking model, the thinking model refers to an implicit knowledge representation model selected by the intelligent cognitive robot for different tasks according to different interactive objects and different interactive tasks. The understanding calculation intelligent unit module is used for emotion calculation, environment state calculation and cognitive target calculation, wherein the emotion calculation refers to calculation of the emotion state of the interactive object according to the obtained emotion information, the environment state calculation refers to comprehensive calculation of the environment related to the current interactive task according to the obtained various environment information, and the cognitive target calculation refers to calculation of the emotion, environment and perception comprehensive condition information related to the target to be achieved according to the target to be achieved by the interactive task. The improved intelligent cognitive robot cognitive module further includes a memory solidification intelligent unit module connected with the cognitive modeling intelligent unit module and the understanding calculation intelligent unit module respectively, the memory solidification intelligent unit module is used for converting the expression information obtained by the use model into a vector respectively.

2. The improved intelligent cognitive robot cognitive module based on ACT-R according to claim 1, wherein, The decision judgment intelligent unit module connected with the reasoning prediction intelligent unit module is further included, the decision judgment intelligent unit module is used for realizing the high-order cognitive ability of the intelligent cognitive robot according to the intention information obtained by the reasoning prediction intelligent unit module.

3. The improved intelligent cognitive robot cognitive module based on ACT-R according to claim 2, wherein, The use model further comprises a feedback learning intelligent unit module connected to the cognitive modeling intelligent unit module, the understanding calculation intelligent unit module and the decision judgment intelligent unit module respectively, and the feedback learning intelligent unit module is used to learn rules and patterns according to feedback information of the cognitive modeling intelligent unit module and the understanding calculation intelligent unit module, and constantly optimize high-order cognitive ability of the feedback learning intelligent unit module.

4. The improved intelligent cognitive robot cognitive module based on ACT-R according to claim 3, wherein, The use model further comprises a comprehensive perception model and an environment model, the comprehensive perception model refers to a perception model selected by the intelligent cognitive robot according to characteristics of perception conditions of the interactive object, and the environment model refers to an environment model provided by the intelligent cognitive robot according to characteristics of an environment where the interactive object is located and an environment required by the interactive task; the feedback learning intelligent unit module can learn rules and patterns according to feedback information of the comprehensive perception model and the environment model, and constantly optimize high-order cognitive ability of the feedback learning intelligent unit module.

5. The improved intelligent cognitive robot cognitive module based on ACT-R according to claim 4, wherein, The comprehensive perception model refers to a perception model selected by the intelligent cognitive robot according to characteristics of perception conditions of the interactive object, and the perception model comprises a dynamic combination of touch perception, hearing perception, electroencephalogram perception, physiological sign perception, force feedback perception and environment perception.

Images