A cognitive decision-making method and system based on multi-dimensional situation coding

Abstract

This invention discloses a cognitive decision-making method and system based on multidimensional situation coding, belonging to the field of artificial intelligence and cognitive computing technology. The method includes: acquiring input data and converting it into an event sequence, thereby generating a six-dimensional external polar field; driving a six-dimensional energy vector to converge from an initial neutral state to a stable state under the action of an energy evolution equation; binarizing the converged energy vector and matching it with a situation coding library to determine the current situation type; generating a structured description and action suggestions for the current situation; selecting a target situation type and planning an evolution path based on situation value weights; and outputting a cognitive decision report. This invention uses events as the native cognitive unit, which can be characters, words, phrases, sentences, or any combination thereof. Through a six-dimensional complete situation space and an energy dynamics emergence mechanism, it provides cognitive systems with endogenous interpretability, a unified situation coding framework, autonomous planning capabilities based on intrinsic value, and online self-evolution capabilities. It can be widely applied to intelligent decision support, complex system situation assessment, and other scenarios.

Description

Technical Field

[0001] This invention belongs to the fields of artificial intelligence, cognitive computing and knowledge representation technology, and specifically relates to a cognitive decision-making method and system based on multidimensional situation coding, which can be used in application scenarios such as intelligent decision support, situation assessment of complex systems, and human-computer interaction cognitive engine. Background Technology

[0002] Current artificial intelligence systems, especially models represented by deep learning, have achieved remarkable results in pattern recognition and prediction tasks. However, their decision-making processes generally suffer from the "black box" problem, lacking interpretability and logical reasoning capabilities, making it difficult to establish user trust in scenarios with high reliability requirements (such as medical diagnosis, industrial control, financial risk control, and robotics). Specifically, existing technologies have the following limitations: (1) Limited cognitive level. Existing systems mostly remain at the level of surface pattern recognition, lacking the ability to understand the deep situation structure and evolutionary laws, and are unable to achieve a complete cognitive cycle from concrete observation to abstract situation assessment, and then from abstract situation to concrete decision-making.

[0003] (2) Insufficient interpretability. The internal representation of deep neural networks is a high-dimensional continuous vector, and the decision-making basis is difficult to present in a way that is understandable to humans. Even if ex post facto interpretation methods are used, it is impossible to restore the true reasoning process of the system.

[0004] (3) Lack of a unified situation coding framework. Different application scenarios have different custom state representation methods, and there is a lack of a complete, interpretable and cross-domain transferable situation classification system.

[0005] (4) Lack of intrinsic value-driven and evolutionary planning capabilities. Existing decision-making systems mostly rely on external reward functions or artificial rules and do not have the ability to autonomously plan evolutionary paths based on intrinsic value assessment.

[0006] (5) Knowledge is fixed and lacks online self-evolution capability. After deployment, the knowledge structure of mainstream models is basically fixed and cannot autonomously adjust its internal representation in continuous interaction with the environment.

[0007] Generative philosophical systems that have been publicly disclosed before the application date (such as *Jishi Jing* and *Jishi Jing Intelligence*, published on the CSDN blog platform) propose the first principle of "genesis is happening" and the core idea of ​​"the sixty-four hexagrams as a complete situational coding space." However, the aforementioned documents are only philosophical descriptions and do not provide specific algorithms or system architectures for implementing this cognitive logic using computer technology. How to transform the abstract idea of ​​situational coding into a computable and executable cognitive decision engine remains a technological gap in this field. Summary of the Invention

[0008] (a) Purpose of the invention The purpose of this invention is to provide a cognitive decision-making method and system based on multidimensional situation coding, so as to solve the problems of poor interpretability, lack of unified situation coding framework, inability to autonomously plan evolution path, and knowledge solidification in existing cognitive systems.

[0009] Specific objectives include: (1) Provides a complete situation coding space, so that any system state can be mapped to a situation type in the space, and the situation type carries a structured semantic interpretation and behavioral guidance.

[0010] (2) Provides an energy dynamics-based situation emergence mechanism that enables the system to autonomously converge to the current situation from the input information, rather than being forced to specify it through rule lookup or classifier.

[0011] (3) Provides a target situation selection and evolution path planning method based on intrinsic value assessment, so that the system can autonomously generate a phased action path that conforms to the law of development of things according to the current situation and intrinsic value weight.

[0012] (4) Provides an online self-evolution mechanism based on event co-occurrence statistics, enabling the system to continuously optimize the mapping relationship between events and situations during operation, and realize the self-growth capability of life-like entities.

[0013] (II) Technical Solution

[0014] To achieve the above objectives, the present invention provides the following technical solution: A cognitive decision-making method based on multidimensional situation coding includes the following steps: Step S1: Obtain the input data and convert it into an event sequence, thereby generating a six-dimensional external polar field.

[0015] The input data includes one or a combination of the following types: - Natural language text, segmented into event sequences using a word segmentation tool; - Structured event sequences, provided directly as a list of event units; - A multidimensional state vector, wherein the dimensions of the state vector include at least one of energy level, knowledge level, resource level, and understanding level.

[0016] Regardless of the initial format of the input data, it is ultimately transformed into an event sequence composed of event units. These event units can be characters, words, phrases, sentences, or any combination thereof. Each event unit corresponds to an independent generation event and carries semantic information that can be mapped to a six-dimensional polar vector. The granularity of the events is determined by the word segmentation tool or the structuring method of the input data.

[0017] The six-dimensional external polar field is a six-dimensional vector, with each dimension taking values ​​in the range of [-1, 1]. Positive values ​​indicate that the polarity of that dimension is biased towards Yang, and negative values ​​indicate that the polarity of that dimension is biased towards Yin.

[0018] When the input data is an event sequence, the system retrieves the six-dimensional polarity vector corresponding to each event by querying a preset innate polarity library. All polarity vectors are then superimposed and multiplied by a preset external field intensity coefficient to obtain the six-dimensional external polarity field. The preset innate polarity library stores the six-dimensional polarity vectors of locked events, which include event units with definite yin-yang attributes. The polarity vectors of the locked events remain unchanged during system operation, providing an unshakable semantic anchor for the entire polarity space. When the input data includes non-locked events not in the innate polarity library, the similarity between the event and each locked event is calculated using a preset semantic similarity model, and a weighted interpolation is used to generate the initial six-dimensional polarity vector for the event.

[0019] When the input data is a multidimensional state vector, each dimension of the state vector is mapped to the corresponding dimension of the six-dimensional external polar field through a preset mapping function.

[0020] Step S2: Input the six-dimensional external polar field into the six-dimensional energy evolution engine to drive the six-dimensional energy vector to converge from the initial neutral state to the stable state, and obtain the converged six-dimensional energy vector.

[0021] The six-dimensional energy vector e = (e1, e2, ..., e6), where ∈[0,1] represents the energy value of the k-th dimension, where 0 represents pure negative and 1 represents pure positive. The initial neutral state has an energy value of 0.5 in each dimension.

[0022] The six-dimensional energy evolution engine updates the energy values ​​of each dimension according to the following energy evolution equation: α + β (ē) ) + γ +η ζ ( 0.5) + in: - These are inter-dimensional coupling terms, calculated based on a pre-defined inter-dimensional coupling matrix, reflecting the interaction between dimensions; - ē represents the global mean of the six-dimensional energy, β (ē) This is a global equilibrium term that drives the energy in each dimension to converge toward the global mean. - For self-sharpening term, = sign( 0.5) Where p > 1, this driving energy accelerates its convergence towards the poles after deviating from the neutral point; - For the perception traction term, the gradient is calculated based on the global cognitive potential U, where U = Var(e) is the variance of the six-dimensional energy; - ζ ( 0.5) represents the source term and the regression term, driving the system to regress to a neutral state when there is no external input; - Let k be the k-th component of the six-dimensional external polar field; - α, β, γ, η, ζ are preset weighting coefficients.

[0023] When a preset convergence condition is met, the evolution stops, and a converged six-dimensional energy vector is obtained. The convergence condition includes: the maximum value of the energy change in each dimension is less than a preset threshold, or the number of evolution steps reaches a preset minimum number of steps and the global cognitive potential U is lower than a preset threshold, or the number of evolution steps reaches a preset maximum number of steps.

[0024] Step S3: Binarize the converged six-dimensional energy vector and match it with a preset situation coding library to determine the current situation type.

[0025] The situation coding library contains 64 basic situation types, each corresponding to a unique six-dimensional binary vector. The binarization rule is: if If the value is greater than or equal to 0.5, then the k-th dimension is 1; otherwise, it is 0.

[0026] The binarized six-dimensional vector is compared one by one with the 64 situation types in the situation coding library, and the situation type with a perfect match or the smallest Hamming distance is determined as the current situation type.

[0027] Step S4: Based on the current situation type, query the preset situation interpretation library and action suggestion library to generate a structured description of the current situation and action suggestions.

[0028] The situation interpretation library pre-stores core situation description text and symbolic meaning description text for each basic situation type. The action suggestion library pre-stores at least one action keyword for each basic situation type.

[0029] Step S5: Based on the current situation type and the preset situation value weight library, select the target situation type and plan the evolution path from the current situation type to the target situation type.

[0030] The situation value weight library pre-assigns a value weight to each basic situation type. A situation type with a higher value weight indicates that the system is more inclined to approach that situation.

[0031] The specific steps for selecting the target situation type are as follows: - Obtain the set of single-step reachable situations for the current situation type in the situation space. The single-step reachable situation refers to a situation type that differs from the six-dimensional binary vector of the current situation type in only one dimension. - For each candidate situation type in the set of single-step reachable situations, calculate the attractiveness value based on its value weight V and the current global cognitive potential U; - Select the candidate situation type with the highest attractiveness value as the target situation type.

[0032] The formula for calculating attractiveness score is: Attr=α (U (1-V))+β V Where V is the value weight of the candidate situation type, U is the current global cognitive situation, and α and β are preset weight coefficients, satisfying α+β= 1.

[0033] The specific steps for planning the evolution path are as follows: taking the current situation type as the starting point and the target situation type as the ending point, the shortest single-step evolution path is searched in the situation space using the A* search algorithm. Each step in the single-step evolution path only changes one dimension of the six-dimensional binary vector.

[0034] Step S6: Output a cognitive decision report.

[0035] The cognitive decision report includes: the identifier, name, and six-dimensional binary vector of the current situation type; the core description and symbolic meaning description of the current situation type; the confidence level of each dimension; the numerical value of the current global cognitive potential; current action suggestions; the identifier, name, and attractiveness value of the target situation type; the evolution path from the current situation type to the target situation type; and target action suggestions.

[0036] Furthermore, the method also includes step S7: online self-evolutionary learning, specifically including: - Record the sequence of input events, current situation type, and six-dimensional energy vector for each cognitive decision-making process; - For each unlocked event in the input event sequence, the six-dimensional polarity vector of the event is fine-tuned according to the six-dimensional binary vector of the current situation type, and the learning rate is modulated by the global cognitive potential; - Maintain the co-occurrence frequency matrix between events. For event pairs that co-occur frequently and have similar polarities, move their six-dimensional polarity vectors closer to each other. - Wherein, the polarity vector of the locking event remains unchanged during fine-tuning and mutual approach.

[0037] Furthermore, the method also includes internal memory and experience reuse: recording the input event sequence, current situation type, target situation type and evolution path of each cognitive decision and storing them in an internal memory bank; when executing step S5 in the subsequent step, retrieving relevant historical experience based on the current situation type and input event sequence, and modulating the attractiveness value of the candidate situation type.

[0038] A cognitive decision-making system based on multidimensional situation coding includes: - Perception mapping module, used to perform step S1; - Energy evolution module, used to perform step S2; - Situation matching module, used to perform step S3; - An interpretation and generation module is used to perform step S4; - Decision planning module, used to execute step S5; - Output module, used to perform step S6.

[0039] Furthermore, it also includes an online learning module for performing step S7.

[0040] Furthermore, it also includes an internal memory module for recording historical experiences and modulating them during decision-making.

[0041] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0042] (III) Beneficial Effects Compared with the prior art, the present invention has the following beneficial effects: (1) Intrinsic interpretability. This invention maps the system state to a complete six-dimensional situation coding space, with each situation type carrying a pre-defined structured semantic description and behavioral guidance. Every step from input events to polarity field generation, to energy evolution situation emergence, and then to target selection and path planning is traceable and auditable, with clear and explicit decision-making basis.

[0043] (2) A unified and complete situation coding framework. This invention adopts a six-dimensional binary situation space (64 basic situation types), which provides a complete situation classification system that can be transferred across domains for any system state, and solves the technical problem of representing various custom states of cognitive systems in different application scenarios.

[0044] (3) Self-organized emergent situation identification. This invention drives the six-dimensional energy to autonomously converge from the initial neutral state to a stable state through the energy evolution equation, rather than through rule lookup or classifier forcibly specifying it. This emergent mechanism makes the situation identification process closer to the essence of biological cognition and has a certain robustness to input noise.

[0045] (4) Autonomous planning based on intrinsic value. The present invention autonomously selects the target situation and plans the evolution path according to the preset situation value weight and the current cognitive certainty, so that the system's decision-making behavior has intrinsic consistency and predictability.

[0046] (5) Online self-evolution capability. This invention enables the system to continuously optimize the mapping relationship between events and states during operation through event polarity fine-tuning and co-occurrence coordination mechanisms, achieving life-like self-growth and self-adaptation capabilities. The learning rate is modulated by the global cognitive potential, ensuring that the system only learns effectively when the cognitive state is stable and reliable.

[0047] (6) Lightweight and edge-side deployment friendly. The core computation of this invention is the iteration of a simple differential equation of a six-dimensional vector, which does not require large-scale matrix operations and GPU acceleration, and can run in real time on resource-constrained edge devices.

[0048] (7) Using events as the primary cognitive unit. This invention uses event sequences as the basic input for cognition. Event units can be characters, words, phrases, sentences, or any combination thereof. This design reflects the philosophical foundation that "events are the basic units of generation"—the system does not limit the granularity of events, but only focuses on whether each event unit can be mapped to a polar vector. The input interface supports multiple data types, all of which are ultimately converted into event sequences, exhibiting good versatility and scalability. Attached Figure Description

[0049] Figure 1 : The overall flowchart of the cognitive decision-making method described in this invention.

[0050] Figure 2 Schematic diagram of the dynamic architecture of the six-dimensional energy evolution engine.

[0051] Figure 3 Example diagram of the evolution path planning from the current situation to the target situation in the situation space.

[0052] Figure 4 Workflow diagram of the online self-evolutionary learning module.

[0053] Figure 5 : Flowchart of the operation of the internal memory module.

[0054] Figure 6 : Module architecture diagram of the cognitive decision-making system described in the present invention. Detailed implementation manners

[0055] It should be understood that the following embodiments are used to illustrate the technical solutions of the present invention rather than to limit them. The core cognitive decision-making process involved is based on the prototype system implemented and verified by the inventor.

[0056] Embodiment 1: Situation emergence and decision-making based on event sequences This embodiment demonstrates how the system processes event sequence inputs. The event units in the event sequence can be of any granularity - single-word events such as "water", "fire", multi-word events such as "artificial intelligence", "yin and yang", and phrase events such as "The weather is nice today".

[0057] Step S1: Obtain the input data and convert it into an event sequence, generating a six-dimensional external polarity field.

[0058] The input is the event sequence: ["water", "fire"], where both "water" and "fire" are single-word events.

[0059] The six-dimensional polarity vectors of the locked events are stored in the preset innate polarity library of the system. For example: - "sky": [1, 1, 1, 1, 1, 1] - "earth": [-1, -1, -1, -1, -1, -1] - "water": [-1, 1, -1, -1, 1, -1] - "fire": [1, -1, 1, 1, -1,​​​​​​​​​​​

[0061] The six-dimensional energy vector is initialized as [0.5, 0.5, 0.5, 0.5, 0.5, 0.5]. Under the drive of the evolution equation, it converges to [1.0, 0.0, 1.0, 0.0, 1.0, 0.0] after about 35 steps, and the global cognitive potential U = 0.25.

[0062] Step S3: Situation matching.

[0063] After binarization, [1, 0, 1, 0, 1, 0] is obtained, which exactly matches the 63rd situation. It is determined that the current situation type is the 63rd situation ("Jiji" situation).

[0064] Step S4: Explanation generation.

[0065] Query the situation explanation library. The core description of the 63rd situation is "The matter has been accomplished, maintain the achieved state and prevent changes", and the action keywords are "maintain the achieved state", "consolidate", and "prevent". The confidence of each dimension is 1.0.

[0066] Step S5: Target situation selection and path planning.

[0067] The single-step reachable situations of the 63rd situation include the 49th situation (flip the fifth dimension). The value weight of the 49th situation is 0.7, and the attractiveness value is calculated as 0.325, which is the highest among all candidate situations. Select the 49th situation as the target situation, and the path is [63, 49].

[0068] Step S6: Output the cognitive decision report.

[0069] System output: The current situation is Jiji (No. 63), the target situation is Ge (No. 49), the evolution path is Jiji → Ge, the current actions are maintain the achieved state, consolidate, and prevent, and the target actions are observe, wait, and follow the trend.

[0070] Example 2: Input processing of events with different granularities This example demonstrates the system's ability to process inputs of events with different granularities.

[0071] Input example 1: Single-word event sequence ["Tian"]. The system queries the innate polarity library and obtains the polarity vector [1, 1, 1, 1, 1, 1] of "Tian", which converges to the 1st situation ("Qian" situation) after energy evolution.

[0072] Input example 2: Multi-word event sequence ["Artificial Intelligence"]. If "Artificial Intelligence" is not in the innate polarity library, the system generates an initial polarity vector through a semantic similarity model and converges to the corresponding situation after energy evolution.

[0073] Input Example 3: Mixed-granularity event sequence ["water", "artificial intelligence", "fire"]. The system acquires or generates the polarity vectors of each event, and superimposes them to generate a six-dimensional external polarity field, driving the energy evolution to converge to the overall situation.

[0074] Input Example 4: Phrase event ["The weather is nice today"]. If the phrase is not in the innate polarity library, the system generates an initial polarity vector through a semantic similarity model, which converges to the corresponding state through energy evolution.

[0075] As seen in the examples above, the system does not restrict the granularity of event units; it only focuses on whether each event unit can be mapped to a six-dimensional polar vector. This flexibility in event granularity allows the system to handle various input formats, ranging from atomic events to composite events.

[0076] Example 3: Emergent Situation and Decision Based on Multidimensional State Vectors This embodiment demonstrates how the system processes abstract state vector inputs.

[0077] Step S1: Acquire the input data and convert it into a six-dimensional external polar field.

[0078] The input is a multidimensional state vector: S = [0.2, 0.8, 0.3, 0.6], representing energy level, knowledge level, resource level, and comprehension level, respectively.

[0079] The system transforms the state vector into a six-dimensional external polar field through a preset mapping function, and then multiplies it by the external field intensity coefficient of 0.4 to obtain the final external polar field.

[0080] Steps S2 to S6: Similar to Example 1, through energy evolution, situation matching, target selection, and path planning, a cognitive decision report is finally output. Under this state combination, the system may emerge as situation number 5 ("need" situation), and the action suggestion is "patience, accumulation, waiting".

[0081] Example 4: Cognition and Learning Including Non-Locked Events This embodiment demonstrates how the system handles inputs containing non-locking events, and the effectiveness of online self-evolutionary learning.

[0082] Initial state: The innate polarity library contains locked events such as "water" and "fire", but the non-locked event "artificial intelligence" is not in the library.

[0083] Initial input: Event sequence ["Artificial Intelligence"]. The system generates an initial polarity vector through a semantic similarity model, which converges to state 30 ("departure" state) through energy evolution.

[0084] Step S7 (Learning): Based on the current convergence situation type, the system fine-tunes the polarity vector of "Artificial Intelligence" to make it move closer to the target polarity direction of situation number 30.

[0085] Second input: Enter "["Artificial Intelligence"]" again. At this point, the polarity vector of "Artificial Intelligence" has been fine-tuned, converging to state 30 more quickly, and the system's understanding of the event is more stable.

[0086] Throughout the learning process, the polarity vectors of locked events such as "water" and "fire" remain unchanged, providing an unshakable semantic anchor for the entire polarity space.

[0087] Example 5: Internal Memory Retrieval Assists Decision Making This embodiment demonstrates how the internal memory module assists in target situation selection.

[0088] Historical experience accumulation: The system has recorded experience: Input ["water", "fire"], current situation number 63, target situation number 49, and obtain positive feedback.

[0089] New input: Event sequence ["water", "fire"]. After the system emerges with the current situation number 63, the internal memory module retrieves matching historical experience and increases the attractiveness value of the target situation number 49 by a small increment, making the system more inclined to choose historically validated effective paths.

[0090] Through this mechanism, the system's decision-making quality gradually improves as its operating time increases.

[0091] Example 6: Implementation of the Perception Layer in Collaboration with a Large Language Model (Predictive Example) This embodiment demonstrates how the perceptual mapping module of the present invention can work in conjunction with a large language model to process complex natural language input. It should be understood that this embodiment describes a foreseeable way in which the technical solution of the present invention is combined with a large language model, and is intended to demonstrate the scalability of the technical solution, not to limit the present invention.

[0092] Step S1-A: Event extraction and polarity initialization assisted by a large language model.

[0093] The system sends the natural language text input by the user to the large language model interface, and calls the large language model to perform the following operations: - Extract key event word sequences from the input text; - Based on common sense knowledge from a large language model, output a preliminary description of the yin-yang polarity for each event word.

[0094] Step S1-B: Polarity calibration and field generation.

[0095] The perception mapping module receives the event sequence and polarity tendency output by the large language model. For each event word: - If the event word is already in the preset innate polarity library, the corresponding locked polarity vector is used directly; - If the event word is not in the innate polarity library, the polarity tendency output by the large language model is used as a reference, and the polarity vectors of each locked event are weighted and interpolated in combination with the semantic similarity model to generate the initial six-dimensional polarity vector of the event.

[0096] The six-dimensional external polar field is obtained by superimposing the polarity vectors of all events and multiplying them by the external field intensity coefficient. Subsequent steps S2 to S7 are the same as in the aforementioned embodiment.

[0097] By collaborating with a large language model, the perception layer of this invention can leverage the powerful natural language understanding capabilities of the large language model to transform user input of any form into structured event polarity data, thus expanding the application scenarios of the system.

Claims

1. A cognitive decision-making method based on multidimensional situation coding, characterized in that, Includes the following steps: - Step S1: Obtain input data and convert it into an event sequence to generate a six-dimensional external polar field; - Step S2: Input the six-dimensional external polar field into the six-dimensional energy evolution engine to drive the six-dimensional energy vector to converge from the initial neutral state to the stable state, and obtain the converged six-dimensional energy vector; - Step S3: Binarize the converged six-dimensional energy vector and match it with a preset situation coding library to determine the current situation type; - Step S4: Based on the current situation type, query the preset situation interpretation library and action suggestion library to generate a structured description of the current situation and action suggestions; - Step S5: Based on the current situation type and the preset situation value weight library, select the target situation type and plan the evolution path from the current situation type to the target situation type; - Step S6: Output a cognitive decision report, which includes the current situation type, a structured description of the current situation, current action recommendations, target situation type, evolution path, and target action recommendations.

2. The method according to claim 1, characterized in that, The input data includes one or a combination of the following types, all of which are ultimately transformed into an event sequence: - Natural language text, segmented into event sequences using a word segmentation tool; - Structured event sequences, provided directly as a list of event units; - A multidimensional state vector, wherein the dimensions of the state vector include at least one of energy level, knowledge level, resource level, and understanding level; The event unit can be a character, word, phrase, sentence, or any combination thereof, and each event unit corresponds to an independent generation event; When the input data is an event sequence, the six-dimensional polar vector corresponding to each event is obtained by querying the preset innate polarity library, and the vectors are superimposed and multiplied by a preset coefficient to obtain the six-dimensional external polarity field. The innate polarity library stores the six-dimensional polarity vectors of locked events, and their polarity vectors remain unchanged during system operation. The initial polarity vectors of non-locked events are generated by weighted interpolation using a semantic similarity model.

3. The method according to claim 1, characterized in that, In step S2, the evolution of the six-dimensional energy vector follows the energy evolution equation, which includes inter-dimensional coupling terms, global equilibrium terms, self-excited sharpening terms, perception traction terms, and source term regression terms. The strength of the self-excited sharpening term increases with the number of evolution steps, and the perception traction term is calculated based on the gradient of the global cognitive potential and introduces a saturation function. The convergence conditions include the energy change being less than a preset threshold, the global cognitive potential being lower than a preset threshold, or the number of evolution steps reaching the upper limit.

4. The method according to claim 1, characterized in that, In step S3, the converged six-dimensional energy vector is binarized according to a threshold and matched with a situation coding library containing 64 basic situation types. In step S5, the set of single-step reachable situations for the current situation type is obtained, the attraction value is calculated based on the situation value weight and the current global cognitive potential, the candidate situation with the largest attraction value is selected as the target situation, and the shortest evolution path is planned using the A* search algorithm.

5. The method according to claim 1, characterized in that, It also includes step S7: online self-evolutionary learning, which specifically includes: recording the input event sequence, current situation type and six-dimensional energy vector for each cognitive decision; fine-tuning the six-dimensional polarity vector of non-locked events, with the learning rate modulated by the global cognitive potential; maintaining the event co-occurrence frequency matrix, and moving the polarity of high-frequency co-occurring and similar event pairs closer together; wherein the polarity vector of locked events remains unchanged.

6. The method according to claim 1, characterized in that, Also includes: Record the sequence of input events, current situation type, target situation type, and evolution path for each cognitive decision and store them in the internal memory bank; In the subsequent execution step S5, the attractiveness value of the candidate situation type is modulated based on the current situation type and relevant historical experience retrieved from the input event sequence.

7. A cognitive decision-making system based on multidimensional situation coding, characterized in that, include: - A perception mapping module, used to perform step S1 as described in claim 1; - Energy evolution module, used to perform step S2 as described in claim 1; - Situation matching module, used to perform step S3 as described in claim 1; - An interpretation and generation module for performing step S4 as described in claim 1; - Decision planning module, used to execute step S5 as described in claim 1; - Output module for performing step S6 as described in claim 1.

8. The system according to claim 7, characterized in that, It also includes an online learning module for performing step S7 as described in claim 5.

9. The system according to claim 7, characterized in that, It also includes an internal memory module for performing the recording, retrieval, and modulation operations as described in claim 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 6.

Images