Brain-like behavior decision calculation model of autonomous robot

Abstract

The invention discloses a brain-like behavior decision calculation model of an autonomous robot, and relates to the field of robots. The method comprises the following steps of: selecting an STM, an LTM, a Up Tree, a Down Tree, Links, Input, Output and Chunkgt; wherein the STM is used for storing and processing short-term information, and the LTM is used for processing information corresponding to different functional modules and simulating division of labor of multiple brain regions; the Up Tree and the Down Tree are uplink and downlink tree information transmission networks; links are dark channels which are established between LTMs in a correlation learning form and are used for direct communication; chunk is an information block for communication transmission between the STM and each LTM, and core communication information is collectively called'brain language '; the Input and the Output are input and output of the system. In a working state, each LTM carries out feedback according to input information, the feedback information is uploaded to the STM through the uplink tree, dynamic migration of information on the STM is realized through competition, an ideological stream is formed, the STM broadcasts the information on the STM to the LTM through the downlink tree, and a loop is formed. According to the invention, collaborative learning can be carried out autonomously, and higher autonomy and intelligent level can be realized.

Description

Technical Field

[0001] The present invention relates to the field of robots, and in particular to a brain-like behavior decision-making computing model of an autonomous robot. Background Art

[0002] At present, robotics technology is developing rapidly, and autonomous robots have been widely used in the fields of industry, medicine, and services, such as autonomous exploration robots in unfamiliar environments and household robots at Stanford University. Although these robots have shown a certain degree of autonomy in specific tasks, they still lack real autonomy, mainly manifested in limited behavioral decision-making capabilities. At present, most robots' behavioral decisions rely on heuristic methods, such as imitation learning and reinforcement learning methods, which enable robots to directly execute the next action or action sequence from their current state according to pre-learned task rules, such as pre-defined rewards or rules in reinforcement learning algorithms and expert example data in imitation learning algorithms. However, when faced with complex and changing actual application environments, these methods often show problems of insufficient generalization ability and low efficiency, that is, pre-learned task rules are difficult to cover the real operating environment of autonomous robots. On the other hand, the recent rapid development of large models has brought new exploration directions. With huge model parameters and huge amounts of data, some researchers have found that large models have an "emergence" phenomenon, that is, after training, new untrained capabilities are stimulated, which seems to be a certain degree of autonomy. However, due to the black box characteristics of neural networks, this "emergence" ability still has huge defects in interpretability and controllability. In addition, existing decision-making algorithms are mostly fixed after being deployed to intelligent agents, and are unable to autonomously optimize and improve their own decision-making capabilities in long-term operation.

[0003] Therefore, technicians in this field are committed to developing a biologically inspired brain-like behavioral decision-making computational model for autonomous robots. Summary of the invention

[0004] In view of the above-mentioned defects of the prior art, the technical problem to be solved by the present invention is how to improve the autonomous behavior capability of the robot.

[0005] To achieve the above objectives, the present invention provides a brain-like behavior decision-making computational model for an autonomous robot, which is described as an octet<STM,LTM,Up Tree,Down Tree,Links,Input,Output,Chunk> , where STM stores and processes short-term information, and LTM processes information corresponding to different functional modules, simulating the division of labor in multiple brain regions; Up Tree and Down Tree are up-down tree information transmission networks; Links are dark channels for direct communication established between the LTMs in the form of associative learning; Chunk is the information block for communication and transmission between the STM and each of the LTMs, and the core communication information is collectively referred to as "brain language"; Input and Output are the input and output of the system; in the working state, each of the LTMs gives feedback based on the input information, and uploads the feedback information to the STM via the up-down tree, and realizes the dynamic migration of the information on the STM through competition, forming a stream of consciousness, and the STM then broadcasts the information on the STM to the LTM via the down-down tree, forming a loop.

[0006] Furthermore, the LTM corresponds to different functional partitions of the robot and realizes different functions. The "perception encoding LTM" realizes acquiring information from the outside world and converting it into "brain language" through encoding; the "information storage and recall LTM" is responsible for converting the input data according to the stored information and outputting the conversion result; the "behavior output LTM" realizes the output of robot action instructions based on the input data; it has great flexibility in the specific implementation form, and multimodal images, that is, text-associated network CLIP can be used for "visual perception" LTM to realize image encoding; the language large model can be used for "storage and recall" LTM to realize the function of calculating and outputting "brain language" from input "brain language".

[0007] Furthermore, the LTM includes a special LTM, which is a world model LTM with recursive characteristics, mainly composed of a learnable predictor, a prediction error comparator and a prediction information storage; the function of the world model LTM is to simulate the process of humans relying on sensory organs to encode information when recognizing the environment.

[0008] Furthermore, unlike directly receiving external environmental information, the world model LTM interacts with the STM short-term memory in a unique way. It indirectly obtains information processed by the "perceptual encoding LTM" and predicts the information on the STM; this prediction is based on the direction of information on the STM "without its own intervention" to evaluate the possible consequences of the current direction of consciousness; the predicted information is then released to the STM, affecting other LTMs and guiding the future direction of consciousness.

[0009] Furthermore, in the prediction of the world model LTM, at any time t, the state of the brain-like behavior decision computing system is expressed as The world model LTM is denoted as W t ,Right now Where e t:0 is all the environmental information up to time t; in addition to evaluating the current direction of consciousness, the world model LTM also predicts itself And through the learning algorithm L (·) Update, that is, W t+1 =L(C t:0 ,e t+1:0 ); this ability distinguishes it from a world model with limited capacity and inability to grow, i.e. It has the potential to continue to grow and evolve; the world model LTM regards the time-varying information on the observed STM as a time series, and is trained in a self-supervised manner to gradually improve its predictive ability and achieve synchronization and consistency with the external environment.

[0010] Furthermore, the "brain language" is expressed as a vector in different LTM modules, and its specific value is obtained by the LTM encoding the input information; there are differences in the encoding methods of different LTMs, so that the same object is encoded by different LTM modules and outputs inconsistent vectors, so a mechanism is needed to achieve brain language information alignment to ensure that different LTM modules can effectively transmit and understand information; one way to achieve brain language alignment is through the "brain language translator" set inside each LTM module, which can be regarded as a vector mapping table, which translates the vector obtained from the STM into the vector corresponding to its own encoder; each LTM obtains the corresponding relationship between the brain language vectors output by different LTMs by comparing the vector of its own brain language output and the vector of other LTM brain language received from the STM, and achieves brain language alignment through collaborative learning to improve environmental adaptability.

[0011] Furthermore, multiple LTMs respond to external inputs simultaneously and transmit outputs to the STM through the ascending tree; the function of the STM is similar to human working memory, which can temporarily store and process a limited amount of information. This limited storage capacity reflects the limitations of the human brain's short-term memory capacity.

[0012] Furthermore, the competition between the STM information is reflected in two aspects: (1) when multiple LTMs send information to the STM through the "uplink and downlink information transmission network", the "uplink and downlink information transmission network" selects the most important information according to the probability of the importance value of each information and submits it to the STM; (2) The STM decides how to select and accept according to the importance value of the information submitted by the LTM, and replaces the existing "brain language" with the selected information to form the "brain-like consciousness" at the current moment and broadcast it to each LTM, thereby forming a continuous "brain language migration" on the STM to form a "stream of consciousness", in which the importance value is provided by the LTM; this decision-making process ensures that the STM always retains the most critical and valuable information.

[0013] Furthermore, when the information on the STM meets specific requirements, the "behavioral output LTM" receives the information and converts it into an output action. In this process, the process in which different LTMs compete to release information and finally output it as an action is called "competitive decision-making". In this way, the brain's ability to prioritize and retain the most important information when faced with information overload is captured. The importance evaluation criteria are provided by the LTM, which requires each LTM to not only be able to generate or process information, but also to evaluate its relative importance to ensure the efficiency and effectiveness of the system.

[0014] Furthermore, in combination with the "brain language translator", the brain language information on the STM is combined with the "natural language encoding and decoding LTM module" to achieve the correspondence between human-understandable semantics and the "brain language" on the STM, thereby achieving readable "brain-like consciousness", that is, the entire information flow in the decision-making process of the model is interpretable. On this basis, artificially defined rules limit and constrain the information on the STM, thereby achieving the controllability of the decision-making process.

[0015] The present invention has the following technical effects:

[0016] (1) Adaptability: The model simulates the cognitive process of the human brain, realizes information competition between different LTMs, effectively integrates and utilizes the functions of each LTM, and enables the robot to autonomously adjust its behavior in a complex dynamic environment.

[0017] (2) Growth: The model achieves autonomous learning and behavior optimization through information exchange and feedback between long and short-term memories (LTM and STM) under unsupervised conditions, enabling the robot to continuously improve the performance of each LTM in long-term operation;

[0018] (3) Interpretability and controllability: The model continuously forms and adjusts the robot’s “brain-like consciousness” by dynamically migrating and releasing information in short-term memory. In addition to providing reference to each LTM through the descending tree, this information flow also makes the decision-making process transparent and visible, and the role of each functional module is clear, which facilitates human intervention for diagnosis, optimization, and constraint, thereby achieving interpretability and controllability.

[0019] The concept, specific structure and technical effects of the present invention will be further described below in conjunction with the accompanying drawings to fully understand the purpose, characteristics and effects of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The overall model structure of a preferred embodiment of the present invention;

[0021] Figure 2 A schematic diagram of the construction of an autoencoder of a perceptual coding LTM according to a preferred embodiment of the present invention;

[0022] Figure 3 A schematic diagram of a world model LTM of a preferred embodiment of the present invention;

[0023] Figure 4 A schematic diagram of brain-speech alignment based on collaborative learning of LTM in a preferred embodiment of the present invention;

[0024] Figure 5 A schematic diagram of an operation mode based on competition decision-making of a preferred embodiment of the present invention;

[0025] Figure 6 The figure is a schematic diagram of the operation mode of a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0026] The following describes several preferred embodiments of the present invention with reference to the drawings in the specification, so that the technical content is clearer and easier to understand. The present invention can be embodied in many different forms of embodiments, and the protection scope of the present invention is not limited to the embodiments mentioned in the text.

[0027] In the drawings, components with the same structure are indicated by the same numerical reference numerals, and components with similar structures or functions are indicated by similar numerical reference numerals. The size and thickness of each component shown in the drawings are arbitrarily shown, and the present invention does not limit the size and thickness of each component. In order to make the illustration clearer, the thickness of the components is appropriately exaggerated in some places in the drawings.

[0028] Aiming at the demand for robot autonomous behavior, an explainable and controllable behavior decision-making computational model is constructed. Based on brain-like information processing mechanisms and cognitive processes, corresponding long-term memory (LTM) modules are established for different robot functional partitions. At the same time, the upper-level processing modules that are performing operations or broadcasting information are modeled as short-term memory (STM) modules, reflecting the dynamic migration of information on STM through competition in the decision-making process of the human brain by multiple functional partitions (modeled as LTM), simulating the formation of continuous consciousness dynamics and the computational model of information fusion and decision-making through attention competition inside the brain. In addition, through the process of multiple functional partitions (LTM) exchanging information through STM and implementing collaborative learning in the decision-making process, the behavior of the LTM module is autonomously optimized under unsupervised conditions, realizing the learning process of the robot in long-term experience and competition feedback.

[0029] Combination Figure 1 The specific content of the brain-like behavior decision-making computing model of an autonomous robot implemented is as follows:

[0030] The computational model of brain-like behavioral decision-making is described by an octet<STM,LTM,Up Tree,Down Tree,Links,Input,Output,Chunk> . Among them, STM is a short-term memory module; LTM is a long-term memory module; Up Tree is an up-tree connecting LTM to STM; Down Tree is a down-tree connecting STM to LTM; Links is a dark channel for direct communication established between LTMs in the form of associative learning; Input obtains environmental information collected by sensors and sends it to the specified LTM processor, and Output transmits the behavior output, that is, the decision result, to the actuator to act on the environment; Chunk is an information block transmitted between STM and each LTM, and the core communication information is collectively referred to as "brain language". Multiple LTMs transmit multiple "brain language" information to STM through the up-tree; STM decides how to replace and publish the information on STM from the information received from several LTMs based on the "competitive decision" mechanism. The published "brain language" is defined as "brain-like consciousness", which is broadcast to each LTM through the down-tree, thus forming a feedback loop and forming a continuous content change on STM.

[0031] Its characteristics are:

[0032] Several LTMs correspond to different functional divisions of the brain / robot and realize different functions. For example: "Perceptual Coding LTM" - obtains information from the outside world and converts it into "brain language" through encoding; "Information Storage and Recall LTM" - converts input data according to the stored information and outputs the conversion result; "Behavior Output LTM" - outputs robot action instructions based on input data. There is great flexibility in the specific implementation form. For example: Multimodal Image-Text Association Network (CLIP) can be used for visual perception LTM to realize image encoding; large language model can be used for "Storage and Recall" LTM to realize the function of calculating output "brain language" from input "brain language", etc. Figure 2 The figure shows a form of "autoencoder" of a perceptual coding LTM. The perceptual input is passed through the encoder to generate the encoding result, that is, the "brain language" output. The encoder can be combined with the decoder through pre-training.

[0033] LTM contains a special kind of LTM, which is a world model LTM with recursive characteristics, such as Figure 3 As shown in the figure, it is mainly composed of a learnable predictor, a prediction error comparator and a prediction information storage. The function of this LTM is to simulate the process of humans relying on sensory organs to encode information when understanding the environment. Unlike directly receiving external environmental information, the "world model LTM" interacts with STM (short-term memory) in a unique way. It indirectly obtains information processed by the "sensory encoding LTM" and predicts the information on the STM. This prediction is based on the direction of information on the STM "without self-intervention" and evaluates the possible consequences of the current direction of consciousness. The prediction information is then released to the STM, affecting other LTMs and guiding the future direction of consciousness. In this process, assuming any time t, the state of the brain-like behavior decision-making computing system can be expressed as The “world model LTM” is denoted by W t .,Right now Where e t:0 is all the environmental information up to time t. In addition to evaluating the current direction of consciousness, the world model LTM also predicts itself And it is updated through the learning algorithm L(·), that is, W t+1 =L(C t:0 ,e t+1:0 ). This ability distinguishes it from a world model with limited capacity and inability to grow, i.e. It has the potential to continue to grow and evolve. The "world model LTM" treats the time-varying information on the observed STM as a time series, and trains it in a self-supervised manner to gradually improve its predictive ability and achieve synchronization and consistency with the external environment.

[0034] Brain language is expressed as vectors in different LTM modules, and its specific value is obtained by LTM encoding the input information. For example, the image encoding LTM assigns a unique vector to each category of image as its brain language. Different LTMs have different encoding methods, which makes the same object output inconsistent vectors after being encoded by different LTMs. Therefore, a mechanism is needed to align brain language information to ensure that different LTM modules can effectively transmit and understand information. One way to achieve brain language alignment is through the "brain language translator" set inside each LTM module. It can be regarded as a vector mapping table, which translates the vector obtained from STM (brain language from other LTMs) into the vector corresponding to its own encoder. Each LTM learns the correspondence between the brain language vectors output by different LTMs by comparing the vector of its own output brain language with the vector of other LTM brain language received from STM, and realizes brain language alignment through collaborative learning to improve environmental adaptability, such as Figure 4 Shown is a schematic diagram of the collaborative learning of audio perception LTM and visual perception LTM.

[0035] Multiple LTMs respond to external inputs simultaneously, and transmit outputs to STM through the ascending tree. The function of STM is similar to human working memory, which can temporarily store and process a limited amount of information. This limited storage capacity reflects the limitations of the human brain's short-term memory capacity. The competition between information is reflected in two main aspects: 1) When multiple LTMs send information to STM through the "uplink and downlink information transmission network", the "uplink and downlink information transmission network" selects the most important information according to the probability of the importance of each information and submits it to STM; 2) STM decides how to select and accept the information based on the importance of the information submitted by LTM, and replaces the existing "brain language" with the selected information to form the "brain-like consciousness" at the current moment and broadcast it to each LTM, such as Figure 5 As shown. This constitutes the continuous "brain language transfer" on STM to form the "stream of consciousness". The importance value is provided by LTM. This decision-making process ensures that STM always retains the most critical and valuable information. When the information on STM meets specific requirements (requirements determined by "behavioral output LTM"), "behavioral output LTM" receives the information and converts it into output actions. In this process, the process in which different LTMs compete to publish information and finally output it as an action is called "competitive decision". In this way, the brain's ability to prioritize and retain the most important information when faced with information overload is captured. The evaluation criteria for importance are provided by LTM, which requires each LTM to not only be able to generate or process information, but also to be able to evaluate its relative importance to ensure the efficiency and effectiveness of the system.

[0036] Combined with the brain language translator, the brain language information on the STM is combined with the "natural language encoding and decoding LTM module" to achieve the correspondence between human-understandable semantics and the "brain language" on the STM, thereby realizing readable "brain-like consciousness", that is, the entire information flow in the decision-making process of the model is explainable. On this basis, artificially defined rules can be used to limit and constrain the information on the STM, thereby achieving controllability of the decision-making process.

[0037] Figure 6 This is a schematic diagram of the operation of the brain-like behavior decision-making computing model, taking the mobile robot's autonomous exploration task in an unfamiliar environment as an example. The single decision-making process is as follows: the visual sensor information will be sent to the "text recognition LTM" and "visual recognition LTM" for processing, and then the two modules will continue to upload the processed information to the STM with a high degree of confidence, such as the brain language encoding of the road sign information and the brain language encoding of the scene visual features; because the confidence of these two pieces of information is high, they will compete successfully to be broadcast on the STM; the "topological map LTM" will receive the brain language broadcast by the STM, determine its own position, and suggest the next moving direction, which is also uploaded to the STM in the form of brain language; at this time, after the competition is successful, the newly broadcast "moving direction suggestion" brain language will be received and processed by the "text recognition LTM" and "visual recognition LTM" again, and the attention will be transferred accordingly to release more accurate recognition information; the "world model LTM" will perform risk / benefit assessment based on the moving direction suggestion and related recognition information decoded from the brain language broadcast by the STM, and upload the assessment results to the STM again; the "motion execution LTM" continuously receives the broadcast brain language, determines the final moving instruction based on the results of multiple rounds of competitive decisions, and outputs it to Gazebo for execution. During this process, dark channels will be formed between LTMs with higher correlation for specific content to accelerate the decision-making process. For example, "text recognition LTM" recognizes the content of the road sign as "No Passage Ahead" multiple times at different times, and the corresponding competitive decision results are all to output the movement instruction of "Stop". In this way, a dark channel will be established between "text recognition LTM" and "world model LTM". When "No Passage Ahead" appears again, the decision-making process will skip the complex "consciousness" transfer and directly pass it to "motor execution LTM" through the dark channel to execute the "stop" action.

[0038] The preferred specific embodiments of the present invention are described in detail above. It should be understood that ordinary technicians in the field can make many modifications and changes based on the concept of the present invention without creative work. Therefore, all technical solutions that can be obtained by technicians in the technical field based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A brain-like behavior decision-making computational model for an autonomous robot, characterized in that: This model is described as an octet<STM,LTM,Up Tree,Down Tree,Links,Input,Output,Chunk> , where STM stores and processes short-term information, and LTM processes information corresponding to different functional modules, simulating the division of labor in multiple brain regions; Up Tree and Down Tree are up-down tree information transmission networks; Links are dark channels for direct communication established between the LTMs in the form of associative learning; Chunk is the information block for communication and transmission between the STM and each of the LTMs, and the core communication information is collectively referred to as "brain language"; Input and Output are the input and output of the system; in the working state, each of the LTMs gives feedback based on the input information, and uploads the feedback information to the STM through the up tree, and realizes the dynamic migration of the information on the STM through competition, forming a stream of consciousness, and the STM then broadcasts the information on the STM to the LTM through the down tree to form a loop.

2. The brain-like behavior decision-making computing model of the autonomous robot according to claim 1, characterized in that: The LTM corresponds to different functional areas of the robot and realizes different functions. The "perception encoding LTM" realizes obtaining information from the outside world and converting it into "brain language" through encoding; the "information storage and recall LTM" is responsible for converting the input data according to the stored information and outputting the conversion result; the "behavior output LTM" realizes the output of robot action instructions based on the input data; it has great flexibility in the specific implementation form, and multimodal images, that is, text-related network CLIP can be used for "visual perception" LTM to realize image encoding; the language large model can be used for "storage and recall" LTM to realize the function of calculating and outputting "brain language" from input "brain language".

3. The brain-like behavior decision-making computing model of the autonomous robot as claimed in claim 2, characterized in that: The LTM includes a special LTM, which is a world model LTM with recursive characteristics, mainly composed of a learnable predictor, a prediction error comparator and a prediction information storage; the function of the world model LTM is to simulate the process of humans relying on sensory organs to encode information when recognizing the environment.

4. The brain-like behavior decision-making computing model of the autonomous robot as claimed in claim 3, characterized in that: Different from directly receiving external environmental information, the world model LTM interacts with the STM short-term memory in a unique way. It indirectly obtains information processed by the "perceptual encoding LTM" and predicts the information on the STM. This prediction is based on the direction of the information on the STM "without self-intervention" and evaluates the possible consequences of the current consciousness direction. The prediction information is then released to the STM, influencing the other LTMs and guiding the future direction of consciousness.

5. The brain-like behavior decision-making computing model of the autonomous robot as claimed in claim 4, characterized in that: In the prediction of the world model LTM, at any time t, the state of the brain-like behavior decision computing system is expressed as The world model LTM is denoted as W t ,Right now Where e t:0 is all the environmental information up to time t; in addition to evaluating the current direction of consciousness, the world model LTM also predicts itself And it is updated through the learning algorithm L(·), that is, W t+1 =L(C t:0 ,e t+1:0 ); this ability distinguishes it from a world model with limited capacity and inability to grow, i.e. It has the potential to continue to grow and evolve; the world model LTM regards the time-varying information on the observed STM as a time series, and is trained in a self-supervised manner to gradually improve its predictive ability and achieve synchronization and consistency with the external environment.

6. The brain-like behavior decision-making computing model of the autonomous robot according to claim 5, characterized in that: The "brain language" is expressed as a vector in different LTM modules, and its specific value is obtained by the LTM encoding the input information; different LTMs have different encoding methods, so that the same object is encoded by different LTM modules and outputs inconsistent vectors. Therefore, a mechanism is needed to align brain language information to ensure that different LTM modules can effectively transmit and understand information; one way to achieve brain language alignment is through the "brain language translator" set inside each LTM module, which can be regarded as a vector mapping table, which translates the vector obtained from the STM into the vector corresponding to its own encoder; each LTM obtains the corresponding relationship between the brain language vectors output by different LTMs by comparing the vector of its own brain language output and the vector of other LTM brain language received from the STM, and realizes brain language alignment through collaborative learning to improve environmental adaptability.

7. The brain-like behavior decision-making computing model of the autonomous robot according to claim 6, characterized in that: The multiple LTMs respond to external inputs simultaneously and transmit the outputs to the STMs through the ascending tree; the function of the STMs is similar to human working memory, which can temporarily store and process a limited amount of information. This limited storage capacity reflects the limitations of the short-term memory capacity of the human brain.

8. The brain-like behavior decision-making computing model of the autonomous robot according to claim 7, characterized in that: The competition between the STM information is reflected in two aspects: (1) when multiple LTMs send information to the STM through the "uplink and downlink information transmission network", the "uplink and downlink information transmission network" selects the most important information according to the probability of the importance value of each information and submits it to the STM; (2) The STM decides how to select and accept the information submitted by the LTM based on its importance value, and replaces the existing "brain language" with the selected information to form the "brain-like consciousness" at the current moment and broadcast it to each LTM, thereby forming a continuous "brain language migration" on the STM to form a "stream of consciousness", in which the importance value is provided by the LTM; this decision-making process ensures that the STM always retains the most critical and valuable information.

9. The brain-like behavior decision-making computing model of the autonomous robot according to claim 8, characterized in that: When the information on the STM meets specific requirements, the "behavior output LTM" receives the information and converts it into an output action. In this process, the process in which different LTMs compete to publish information and finally output it as an action is called "competitive decision"; In this way, the brain's ability to prioritize and retain the most important information when faced with information overload is captured; the importance assessment criteria are provided by the LTM, which requires that each LTM not only be able to generate or process information, but also be able to assess its relative importance to ensure the efficiency and effectiveness of the system.

10. The brain-like behavior decision-making computing model of the autonomous robot according to claim 9, characterized in that: Combined with the "brain language translator", the brain language information on the STM is combined with the "natural language encoding and decoding LTM module" to achieve the correspondence between human-understandable semantics and the "brain language" on the STM, thereby achieving readable "brain-like consciousness", that is, the entire information flow in the decision-making process of the model is interpretable. On this basis, artificially defined rules are used to limit and constrain the information on the STM, thereby achieving controllability of the decision-making process.

Images