An evolutionary method based on rigid constraints and rule adaptation

Abstract

This invention discloses a rule-adaptive evolutionary method based on rigid constraints, relating to the fields of artificial intelligence and information processing technology. The method includes the following steps: collecting output score records generated during the continuous operation of an intelligent agent, organizing them chronologically to form a score change sequence, and marking repeatedly occurring score stagnation positions within the score change sequence to form score stagnation references. By constructing a score change sequence and setting score stagnation references and rule stagnation segments, this invention establishes a correspondence between the score change process and the rule's position, thereby achieving accurate positioning of the rule adjustment location. Simultaneously, by forming candidate rule adjustment segments and a rule adjustment progression order, the rule content changes can proceed segment by segment along the score change path, and after the score stabilizes, the rule content remains unchanged, thus achieving orderly adaptive evolution of the rule system within rigid constraints.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and information processing technology, and specifically to an evolutionary method based on rule-adaptive rigid constraints. Background Technology

[0002] Rigid constraint-based adaptive evolution of rules refers to the process of structurally expressing rules, originally in natural language, during the operation of an intelligent agent. Each rule is labeled with a different rigidity level, and the importance of rules is managed hierarchically. While ensuring that core business bottom-line rules remain unchanged, adjustable rules are dynamically optimized and iteratively updated. When the agent's output fails to meet the target requirements, the system analyzes the scoring gap, historical iterations, and the current rule state to determine the source of the problem and generates optimization suggestions for adjustable rules. Semantic adjustments are made to the rule content while strictly adhering to rigidity level constraints, resulting in a new rule version. This process, through continuous iteration and verification, enables the rules to be gradually optimized and evolved within safe boundaries, allowing the rule system to continuously adapt to new needs and environmental changes while maintaining the stability of core constraints.

[0003] The existing technology has the following shortcomings: In existing technologies, rule-based adaptive evolution based on rigid constraints typically relies on feedback from the agent's output to optimize rules. However, this approach is prone to several problems in practice. First, when the agent's output fails to meet expectations, the system often needs to determine whether the problem stems from unreasonable rule settings or insufficient model execution capability. However, existing technologies lack an effective differentiation mechanism, easily misjudging execution-level issues as rule problems, triggering unnecessary rule adjustments and causing the optimization direction to deviate. Second, as the number of rules increases and their structure becomes more complex, implicit relationships or conflicts easily form between different rules. Without a unified analysis and coordination mechanism, rule adjustments can easily trigger a chain reaction, reducing the overall stability of the rule system. Furthermore, existing technologies often employ overwrite updates for rule modification, lacking a complete version record and rollback mechanism. Once the optimized rule's effectiveness declines, it becomes difficult to promptly revert to a stable version, thus affecting the reliability of the system's continued operation.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a rule-adaptive evolutionary method based on rigid constraints to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a rule-adaptive evolutionary method based on rigid constraints, comprising the following steps: The system collects output score records generated during the continuous operation of the intelligent agent and organizes them into a score change sequence according to time. The system marks the positions where the score repeatedly stops in the score change sequence to form a score stop reference, so as to establish the correspondence between the position of rule action and score change. By reviewing the time segments corresponding to the scoring change sequence based on the scoring stagnation reference, the order in which each rule participated in the scoring within the time segment is organized segment by segment. The positions where the score continues to decline while the rule content remains unchanged are identified and marked in the scoring change sequence to form rule stagnation segments, so as to present the correspondence between scoring stagnation and rule content. Based on the rule stagnation section, continue to expand the score change sequence, read the score change process before and after the rule stagnation section, sort out the score change rhythm within the rule stagnation section segment by segment, extract the positional relationship where the score has changed and the rule content remains unchanged, and form a rule adjustment candidate section in the score change sequence. By reviewing the time range corresponding to the score change sequence of the candidate segments for rule adjustment, the changes in rule content within the candidate segments are sorted out segment by segment, and the order of rule adjustment is rearranged according to the score change position corresponding to the rule stagnation segment, forming the order of rule adjustment progress, so as to limit the scope of subsequent rule content changes. Based on the rule adjustment and advancement sequence, continue to read the subsequent changes in the score change sequence. Adjust the rule content segment by segment in the time interval corresponding to the rule adjustment and advancement sequence, and keep the rule content unchanged after the score change returns to stability, so as to form a rule adaptive evolution process that is gradually advanced within the rigid constraint.

[0007] Preferably, the steps for collecting output score records and forming score change sequences during the continuous operation of the intelligent agent are as follows: The system collects output score records generated during the continuous operation of the intelligent agent and arranges these records according to the order of score generation time to form a score change sequence. Based on the order of the score records in the score change sequence, the score value change status is sorted out segment by segment. The positions where the score value remains consistent in multiple consecutive positions are identified in the score change sequence and marked to form the positions where the score repeatedly stays. Based on the number of the repeated points of scoring in the scoring change sequence, the points of repeated scoring are centrally organized to form a scoring reference. Based on the reference position of the scoring stop, the time segments corresponding to the scoring change sequence are sorted out segment by segment, and the participation of the rule in the scoring in the corresponding time segment is recorded to form a correspondence between the position of the rule and the scoring change.

[0008] Preferably, the steps for reviewing the time intervals corresponding to the score change sequence around the score stagnation reference and forming regular stagnation intervals are as follows: The scoring change sequence is reviewed and organized around the corresponding position of the scoring stop reference in the scoring change sequence, and the scoring records before and after the scoring change sequence are continuously expanded according to the scoring stop reference position to form the time segment corresponding to the scoring stop reference. Based on the corresponding time intervals of the scoring reference, the participation of the corresponding rules in the scoring change sequence is sorted out segment by segment, and the order of participation of each rule in the scoring is arranged according to the time sequence of the scoring change sequence to form a record of the order of participation of the rules in the scoring. Based on the rules, the scoring order records are used to jointly organize the scoring change trend and rule content change status in the scoring change sequence, and the positional relationship of continuous score decline and unchanged rule content is identified in the scoring change sequence. Based on the positional relationship of continuously decreasing scores while the rule content remains unchanged, the consecutive positions corresponding to the score change sequence are marked and organized to form rule stagnation segments and present the correspondence between score stagnation and rule content.

[0009] Preferably, based on the corresponding position of the rule stagnation segment in the score change sequence, the positions in the score change sequence that continuously decrease in score are continuously organized, and the rule content state corresponding to the position of continuous score decrease is recorded segment by segment to form a correspondence record between the state of continuous score decrease and the state of unchanged rule content.

[0010] Preferably, the steps for expanding the score change sequence based on the rule stagnation segment and forming the rule adjustment candidate segment are as follows: The scoring change sequence is continuously expanded around the corresponding position of the rule stagnation section in the scoring change sequence, and the scoring records before and after the rule stagnation section are sequentially arranged to form a continuous scoring change section. Based on the continuous range of score changes, the score records within the rule stagnation range are organized segment by segment to form a record of the rhythm of score changes within the rule stagnation range. Based on the record of the rhythm of score changes within the rule stagnation zone, the score value change state and rule content state in the score change sequence are jointly organized, and the positional relationship of score changes and rule content unchanged is identified in the score change sequence. Based on the positional relationship where the scores have changed but the rule content remains unchanged, the corresponding consecutive positions in the score change sequence are sorted sequentially to form candidate segments for rule adjustment in the score change sequence.

[0011] Preferably, the steps for reviewing the score change sequence of candidate segments for rule adjustment and forming the rule adjustment implementation order are as follows: The scoring change sequence is reviewed and organized around the corresponding position of the candidate segment for rule adjustment in the scoring change sequence. Then, the scoring records before and after the candidate segment for rule adjustment are continuously expanded according to the corresponding position of the candidate segment for rule adjustment to form the time range corresponding to the candidate segment for rule adjustment. Based on the rules, the time range corresponding to the candidate segments is adjusted, and the changes in the rule content corresponding to the score records in the score change sequence are sorted out segment by segment, and the rule content change records are formed according to the time order of the score change sequence. Based on the record of rule content changes and the corresponding score change positions in the rule stagnation section, the rule participation positions in the candidate rule adjustment section are organized accordingly, and the rule adjustment order is arranged according to the order of the score change positions in the rule stagnation section. The order of rule adjustment is determined by arranging the rules and then the corresponding positions of the score change sequence are sorted out segment by segment according to the rule adjustment order to limit the scope of rule content change.

[0012] Preferably, the corresponding positions of the scoring change sequence are continuously organized according to the rule adjustment and advancement order, and the scoring change status in the scoring change sequence is recorded in correspondence with the rule content change status, forming a rule content change path in the scoring change sequence, so as to limit the rule content change process to advance sequentially along the scoring change path.

[0013] Preferably, the process of adjusting the order of advancement according to the rules to continue reading the subsequent changes in the score change sequence and forming a rule-adaptive evolutionary process is as follows: Based on the order of rule adjustments, the subsequent score records in the score change sequence are continuously read at the corresponding positions in the score change sequence to form the subsequent change process of the score change sequence; By combining the subsequent changes in the scoring change sequence with the corresponding time periods of the rule adjustment implementation order, the rule content at the corresponding position in the scoring change sequence is sorted out segment by segment, and the rule content is adjusted segment by segment according to the rule adjustment implementation order. The process of adjusting the rules segment by segment continuously organizes the status of score value changes in the score change sequence, and identifies the stable position of score change in the score change sequence. By combining the score change to restore the stable position, the rule content at the corresponding position in the score change sequence is maintained and organized, and the rule adaptive evolution process is gradually promoted within the rigid constraint range formed by the rule adjustment and advancement order.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention constructs a scoring change sequence and sets scoring stagnation references and rule stagnation sections within it, establishing a corresponding relationship between the scoring change process and the position where the rule applies. This allows for the identification of positions where the score continuously decreases while the rule content remains unchanged during the scoring change process. In this way, the scoring change trend in the scoring change sequence and the state of the rule content can be clearly mapped, so that the rule content change process no longer relies on experience-based judgment but is located and organized based on the scoring change process. This provides a clear reference for rule adjustment, reduces blind changes during rule adjustment, and allows rule content changes to unfold gradually around the scoring change state.

[0015] This invention generates candidate segments for rule adjustments within a score change sequence and further generates a rule adjustment progression order. This allows rule content changes to proceed segment by segment according to the time intervals defined by the score change sequence. Once the score changes stabilize, the rule content remains unchanged, thus keeping rule content changes within rigid constraints. In this way, the rule content change process forms a continuous progression path, enabling the rule adaptive evolution process to unfold gradually with reference to the score change sequence. This maintains the stability of the rule system during its evolution and allows the rule adjustment process to continuously optimize rule content within a defined progression range. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart of a rule-adaptive evolutionary method based on rigid constraints according to the present invention.

[0018] Figure 2 This is a three-level hierarchical diagram of rigid constraints in this invention.

[0019] Figure 3 This invention relates to the role of rigidity levels in rule optimization.

[0020] Figure 4 This is the structure of the strategy analyzer of the present invention.

[0021] Figure 5 This is a flowchart of the semantic-level optimization process for the rules in this invention.

[0022] Figure 6 The flowchart for optimizing the boundary control rules of this invention is shown.

[0023] Figure 7This is a genealogical tracing diagram of the rule version of this invention.

[0024] Figure 8 This is a diagram of the conditional rule persistence mechanism of the present invention.

[0025] Figure 9 This is a diagram of the overall system architecture of the present invention. Detailed Implementation

[0026] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0027] This invention provides, for example Figure 1 The illustrated rule-adaptive evolutionary method based on rigid constraints includes the following steps: The system collects output score records generated during the continuous operation of the intelligent agent and organizes them into a score change sequence according to time. The system marks the positions where the score repeatedly stops in the score change sequence to form a score stop reference, so as to establish the correspondence between the position of rule action and score change. As the agent continuously operates and generates output results, the output score records are continuously organized and processed to gradually construct a time-series structure that reflects score changes. Within this sequence structure, the positions where scores remain are identified to form a score reference, laying the foundation for subsequent identification of the correspondence between the positions where rules apply and score changes. The specific implementation steps are as follows: Output score records generated during the continuous operation of the intelligent agent are collected uniformly and organized in a unified chronological order to form a complete score change sequence. During the agent's task execution, each output result corresponds to an output score record, which includes the score value and the time point of score generation. By collecting all output score records generated during the continuous operation phase and arranging them in chronological order of score generation, a score change sequence unfolds sequentially. In this score change sequence, each score record is arranged chronologically, allowing the sequence to fully represent the process of score value changes over time during continuous operation. Simultaneously, during the chronological organization of the output score records, each score record is numbered and identified within the score change sequence, ensuring each record has a unique positional identifier. This method not only reflects the continuous state of score value changes over time but also provides a clear sequence basis for further marking and locating score change positions.

[0028] Based on the established scoring change sequence, the scoring changes within the sequence are continuously observed, and locations where scores repeatedly remain are identified. In the scoring change sequence, scores may rise, fall, or remain unchanged at different time points. When a score remains unchanged across multiple consecutive positions or repeatedly remains unchanged between adjacent scores, this area constitutes a location where scores repeatedly remain unchanged. Following the established order of the scoring records in the scoring change sequence, the score changes for each record are segmented, and regions where scores remain consistent across multiple consecutive positions are identified. When a score remains unchanged across multiple consecutive records or repeatedly remains unchanged between adjacent records, the corresponding location is marked as a location where scores repeatedly remain unchanged, and this location is explicitly marked in the scoring change sequence. In this way, multiple locations where scores repeatedly remain unchanged can be identified in the scoring change sequence. This allows the scoring change sequence to not only reflect the trend of scoring changes but also the stable states of scores at certain stages, providing a basis for further establishing reference points for score dwell.

[0029] Based on the identified recurring points of scoring, these points in the scoring change sequence are systematically organized and used to form a scoring dwell reference. Specifically, the marked recurring points in the scoring change sequence can be considered key observation points in the scoring change process. By centrally organizing each recurring point in the scoring change sequence and recording its position number, a scoring dwell reference describing the scoring dwell state can be formed. In the process of forming the scoring dwell reference, each recurring point corresponds to a scoring dwell reference position, which includes the position number in the scoring change sequence and the corresponding scoring value. This allows the scoring dwell reference to fully reflect the dwell position and state of the scoring value in the scoring change sequence. Simultaneously, during the formation of the scoring dwell reference, the sequential relationship of the scoring changes before and after the reference in the scoring change sequence is also organized, enabling the scoring dwell reference to reflect the change path of the scoring value before and after dwelling in the scoring change sequence. The scoring dwell reference formed in this way not only reflects stable dwell positions in the scoring change sequence but also the continuous change relationship between various dwell positions.

[0030] Based on the established scoring dwell references, these references are mapped to the positions where rules apply, thus establishing a correspondence between these positions and scoring changes. During agent operation, each rule participates in the scoring process at a specific time stage; therefore, the positions where rules apply can be mapped to time segments within the scoring change sequence. Around the established scoring dwell reference positions, the time segments corresponding to these reference positions in the scoring change sequence are organized segment by segment, and the rule's effect within each time segment is recorded segment by segment. By mapping the scoring dwell reference positions to the positions where rules apply segment by segment, a correspondence between the positions where rules apply and scoring changes is established. In this correspondence, each scoring dwell reference position can be associated with a rule position within the scoring change sequence, enabling a corresponding association between the scoring dwell state and the rule position within the scoring change sequence. The correspondence between the rule's position and the score change established in this way can provide a basis for further analysis of the relationship between score change and rule content. This allows the score change sequence to play a basic reference role in the subsequent identification of rule stagnation segments and rule adjustment candidate segments, thus enabling the subsequent rule adaptive evolution process to be based on a clear correspondence between the score change sequence and the rule's position.

[0031] By reviewing the time segments corresponding to the scoring change sequence based on the scoring stagnation reference, the order in which each rule participated in the scoring within the time segment is organized segment by segment. The positions where the score continues to decline while the rule content remains unchanged are identified and marked in the scoring change sequence to form rule stagnation segments, so as to present the correspondence between scoring stagnation and rule content. Based on the established scoring change sequence and scoring stagnation reference, the time segments in the scoring change sequence are reviewed and organized around the positions corresponding to the scoring stagnation references. By continuously analyzing the order in which each rule participates in the scoring, a correspondence is established between the scoring change process and the rule content change status. This identifies positions where the score continuously declines while the rule content remains unchanged, and these positions are marked in the scoring change sequence to form rule stagnation segments, thus presenting the correspondence between scoring stagnation and rule content. The specific implementation steps are as follows: Based on the established rating pause reference markings in the rating change sequence, the sequence is reviewed and organized around the corresponding positions of these references to determine the corresponding time segments. Specifically, each rating pause reference corresponds to a location where ratings repeatedly pause, with each reference position having a distinct position number within the sequence. Around each reference position, the rating records before and after that position are continuously expanded, creating a continuous observation segment of rating changes near that position. Within this observation segment, the sequence is organized chronologically, ensuring a complete time segment correspondence between the reference positions and the preceding and following rating changes. This method creates several time segments around the reference positions, allowing the references to not only indicate the location of the pauses but also link them to specific time segments within the sequence, providing a temporal basis for subsequently identifying the relationship between rating changes and rule participation.

[0032] With the time interval corresponding to the scoring reference already determined, the order in which each rule participates in the scoring within that time interval is systematically organized segment by segment, ensuring a continuous correspondence between the scoring change process and the rule participation process. Specifically, within the time interval corresponding to the scoring reference, the agent participates in the scoring process according to different rules at different time points. Therefore, each scoring record position in the scoring change sequence can be associated with the rule participating in the scoring at that time. Around the time interval corresponding to the scoring reference, the rule participation situation corresponding to each scoring record position in the scoring change sequence is systematically organized segment by segment, and the order in which each rule participates in the scoring is arranged according to the chronological order of the scoring records in the scoring change sequence. During this process, the position of each rule participating in the scoring is marked with a corresponding position in the scoring record in the scoring change sequence, thus forming a rule participation scoring order record. This ensures that the scoring records in the scoring change sequence not only reflect changes in scoring values ​​but also the order in which the rules participate in the scoring process. Through this segment-by-segment organization method, a rule participation scoring order sequence can be formed within the time interval corresponding to the scoring reference, ensuring a complete correspondence between the scoring change process and the rule participation scoring process.

[0033] Based on the established rule-based scoring sequence, the scoring trend and rule content changes within the scoring sequence are jointly analyzed to identify positions where the score continuously decreases while the rule content remains unchanged. Specifically, within the time interval corresponding to the scoring reference, the score values ​​in the scoring sequence change as the agent continuously operates, with some score record positions showing a continuous downward trend. Around the rule-based scoring sequence, positions where consecutive score decreases are analyzed segment by segment, while simultaneously recording the rule content changes at the corresponding positions. During this analysis, the rule content corresponding to each score record position in the scoring sequence is continuously expanded. When the score value decreases at multiple consecutive positions within a certain time interval, while the rule content at the corresponding positions remains unchanged, this positional relationship is identified as a position where the score continuously decreases while the rule content remains unchanged. In this way, several positional relationships where the score continuously decreases while the rule content remains unchanged can be identified within the scoring sequence, establishing a clear correspondence between the scoring trend and the rule content state.

[0034] Based on identifying the positional relationship where scores continuously decrease while rule content remains unchanged, this positional relationship is marked and organized within the score change sequence to form rule stagnation segments, thus demonstrating the correspondence between score stagnation and rule content. Specifically, around the identified positional relationship where scores continuously decrease while rule content remains unchanged, consecutive positions in the score change sequence are uniformly marked, and these consecutive positions are organized into rule stagnation segments. Within each rule stagnation segment, each position simultaneously possesses two characteristics: first, the score value in the score change sequence shows a continuous decrease; second, the rule content at the corresponding position remains unchanged. By centrally marking positions in the score change sequence that meet these characteristics, several rule stagnation segments can be formed. Each rule stagnation segment can reflect the correspondence between the score change state and the rule content state in the score change sequence, enabling a corresponding mapping between the continuous decrease in scores and the unchanged state of rule content. The rule stagnation segments formed in this way can not only show the correspondence between scoring stagnation and rule content, but also provide a basic positional reference for further sorting out the scoring change sequence around the rule stagnation segment, extracting candidate segments for rule adjustment, and determining the order of rule adjustment. This allows the scoring change sequence to continue to play a reference role in the subsequent adaptive evolution of rules.

[0035] Based on the rule stagnation section, continue to expand the score change sequence, read the score change process before and after the rule stagnation section, sort out the score change rhythm within the rule stagnation section segment by segment, extract the positional relationship where the score has changed and the rule content remains unchanged, and form a rule adjustment candidate section in the score change sequence. Having identified rule stagnation sections in the scoring change sequence, the scoring change sequence continues to revolve around these stagnation sections. By continuously organizing the scoring changes before and after these stagnation sections, a clearer correspondence is established between the rhythm of scoring changes in the scoring change sequence and the state of rule content changes. Based on this, the positional relationships where the scoring has changed but the rule content remains unchanged are extracted, thus forming candidate segments for rule adjustment in the scoring change sequence. The specific implementation steps are as follows: Based on the established stagnant sections in the scoring change sequence, the sequence is further expanded, and the scoring changes before and after these stagnant sections are continuously read and sequentially organized. Specifically, each stagnant section corresponds to a set of consecutive position numbers in the scoring change sequence, indicating a continuous decrease in scores while the rule content remains unchanged. Around these position numbers, the scoring records before and after the stagnant section are continuously expanded, forming a complete continuous segment with the stagnant section and its preceding and following positions. During this organization, the stagnant section is used as a central reference point in the scoring change sequence, and the scoring records before and after it are read progressively around this central reference point, creating a continuously expanding segment of scoring changes near the stagnant section. In this way, the stagnant section and its preceding and following scoring changes form a complete and continuous observation range within the scoring change sequence, providing a foundation for subsequently identifying the rhythm of scoring changes.

[0036] Having completed the continuous unfolding of the scoring changes before and after the rule-stagnant section, the scoring change rhythm within the rule-stagnant section is systematically organized segment by segment, ensuring a complete presentation of the scoring change rhythm within the scoring change sequence. Specifically, within the rule-stagnant section, the scoring records in the scoring change sequence are arranged sequentially according to time, with each record corresponding to a scoring value and a corresponding rule content state. Around the scoring records within the rule-stagnant section, the changing states of the scoring values ​​in the scoring change sequence are systematically organized segment by segment, creating a clear trajectory of scoring change rhythm within the rule-stagnant section. During this process, the scoring records within the rule-stagnant section are divided into several consecutive positional segments according to time, and the changes in the scoring values ​​within each segment are recorded, thus forming a record of the scoring change rhythm within the rule-stagnant section. This segment-by-segment organization method allows for a complete presentation of the scoring change rhythm within the rule-stagnant section, establishing a clear correspondence between the scoring change rhythm and the positional relationship of the rule-stagnant section.

[0037] Based on the unfolded score change sequence and the formation of continuous score records in the rule-stagnant section, the change amplitude between adjacent score records in the score change sequence is cumulatively expressed to construct a score change driving index for characterizing the rhythm of score changes, which is specifically described by the following formula: In the formula, It is a score change driving index, used to represent the overall fluctuation intensity of score changes within the rule stagnation zone; It is the [number]th in the score change sequence The rating value corresponding to each rating record It is the total number of scoring records within the rule-free stagnation zone; This represents the magnitude of score change between adjacent score records. This formula allows for a quantitative expression of the rhythm of score change within a stagnant section of the rule. If the score change-driving index continues to increase while the rule content remains unchanged, it indicates that score change is still occurring, and that the change is not caused by rule adjustment but rather stems from changes in the execution process or external input. Therefore, this score change-driving index can serve as an important basis for judging the rhythm of score change, establishing a measurable correspondence between the rhythm of score change and the state of the rule content. Based on this index, regions with significant score changes but unchanged rules can be further screened, providing quantitative support for subsequent extraction of candidate sections for rule adjustment. This allows the identification of candidate sections to move beyond relying on single-location judgments and instead be based on a holistic analysis of the continuous rhythm of score change.

[0038] Having systematically organized the rhythm of score changes within the stagnant rule segments, the scoring change states and rule content states in the scoring change sequence are jointly analyzed to extract positional relationships where scores have changed while rule content remains unchanged. Specifically, within the stagnant rule segments and in consecutive scoring change segments before and after them, the score value changes and rule content states corresponding to each scoring record position are continuously observed. When the score value changes at certain positions in the scoring change sequence, while the rule content at the corresponding positions remains unchanged, this positional relationship can be identified as one where the score has changed while the rule content remains unchanged. During the analysis, positions in the scoring change sequence that meet the above conditions are marked segment by segment, and their position numbers in the scoring change sequence are recorded, establishing a correspondence between the scoring change states and rule content states. In this way, several positional relationships where scores have changed while rule content remains unchanged can be extracted from the scoring change sequence, providing a foundation for subsequently forming candidate segments for rule adjustments.

[0039] Based on quantifying the rhythm of rating changes and identifying rating change behaviors, the rating change state and rule content state in the rating change sequence are jointly characterized to extract the positional relationship where the rating has changed but the rule content remains unchanged. This is specifically described by the following formula: In the formula, It is the static deviation of the rule, used to represent the degree of deviation between a change in the score and an unchanged rule content. It is an indicator of score change, when the first The value is 1 when the rating record changes, and 0 when the rating does not change. It is the rule content stability factor, when the first... The value is 1 when the rating record remains unchanged and 0 when the rule changes. This formula allows for a joint expression of the relationship between rating changes and rule content. When both the rating change indicator and the rule content stability factor remain 1 within a continuous segment, the static deviation of the rule will accumulate within that segment, forming a clear deviation feature. This deviation feature accurately reflects the positional relationship where rating changes have occurred but the rule content has not. Therefore, a set of positions continuously satisfying this condition can be extracted from the rating change sequence, and this set can be organized to form candidate segments for rule adjustment. By introducing the static deviation of the rule, quantitative identification of candidate segments can be achieved, providing a clear basis for the formation of candidate segments for rule adjustment, thereby enhancing the positioning accuracy and stability during the adaptive evolution process of the rule.

[0040] Based on the extracted positional relationships within the score change sequence where scores have changed while rule content remains unchanged, these positional relationships are continuously organized within the score change sequence to form rule adjustment candidate segments. Specifically, around the extracted score change positions, the relationships between adjacent positions in the score change sequence are sequentially organized. When multiple consecutive positions satisfy the condition that scores have changed while rule content remains unchanged, these consecutive positions are organized into a rule adjustment candidate segment. Within the rule adjustment candidate segment, each position reflects a correspondence between the score change state and the rule content state; that is, the score value in the score change sequence has changed, while the rule content remains unchanged. By centrally organizing these positional relationships, several rule adjustment candidate segments can be formed within the score change sequence, ensuring a correspondence between the score change process and the maintained rule content state. These rule adjustment candidate segments, formed in this way, provide a foundation for subsequently reviewing the corresponding time range of the score change sequence around these candidate segments and rearranging the rule adjustment order. This allows the rule adaptive evolution process to gradually advance based on a clear correspondence between the score change sequence and the rule content state.

[0041] By reviewing the time range corresponding to the score change sequence of the candidate segments for rule adjustment, the changes in rule content within the candidate segments are sorted out segment by segment, and the order of rule adjustment is rearranged according to the score change position corresponding to the rule stagnation segment, forming the order of rule adjustment progress, so as to limit the scope of subsequent rule content changes. Based on the candidate segments for rule adjustments already identified in the score change sequence, the process continues by reviewing the corresponding time range of the score change sequence around these candidate segments. By systematically analyzing the rule content changes within each candidate segment and considering the position of the rule stagnation segments within the score change sequence, the order of rule adjustments is rearranged. This creates a sequence for rule adjustment implementation, limiting the scope of subsequent rule content changes and ensuring that rule content changes unfold systematically within the constraints of the score change sequence. The specific implementation steps are as follows: Based on the identification of candidate segments for rule adjustments within the score change sequence, the corresponding time range of the score change sequence is reviewed around these candidate segments to fully unfold their positional relationship within the score change sequence. Specifically, each candidate segment corresponds to a set of consecutive position numbers in the score change sequence, indicating a positional relationship where the score has changed while the rule content remains unchanged. Around these position numbers, the score records before and after the candidate segment in the score change sequence are continuously unfolded, allowing the candidate segment to form a continuous time range with its preceding and following score changes. During this unfolding process, using the start and end positions of the candidate segment in the score change sequence as central references, adjacent score records are sequentially unfolded, creating a complete time range around the candidate segment. This method establishes a complete correspondence between the candidate segment and the score change process within the score change sequence, providing a foundational time range for subsequent segment-by-segment analysis of rule content changes within the candidate segment.

[0042] Having completed the expansion of the time range corresponding to the candidate segments for rule adjustments, the changes in rule content within each candidate segment are systematically organized segment by segment, ensuring a correspondence between the score change status and the rule content change status in the score change sequence. Specifically, within the time range corresponding to the candidate segments for rule adjustments, each score record position in the score change sequence corresponds to a score value and a corresponding rule content status. Around each score record position within the candidate segments for rule adjustments, the rule content corresponding to each position is continuously organized, and the changes in rule content are recorded segment by segment according to the chronological order of the score change sequence. During this organization process, the score records within the candidate segments for rule adjustments are divided into consecutive positions, and the change status of the rule content in each consecutive position is recorded, ensuring a complete representation of the rule content's trajectory in the score change sequence. This segment-by-segment organization method allows the changes in rule content within the candidate segments for rule adjustments to form a continuous record in the score change sequence, thus providing foundational data for subsequently rearranging the rule adjustment order based on the score change positions corresponding to rule stagnation segments.

[0043] Based on the segmented analysis of rule content changes within the candidate rule adjustment sections, and considering the position of rule stagnation sections within the score change sequence, the rule adjustment order is rearranged to ensure a correspondence between the order of rule content changes and the score change process. Specifically, rule stagnation sections are marked in the score change sequence as positions where scores continuously decrease while rule content remains unchanged; therefore, the score change positions corresponding to rule stagnation sections serve as important references for rearranging the rule adjustment order. Around the score change positions corresponding to rule stagnation sections, the participation positions of each rule within the candidate rule adjustment section in the score change sequence are analyzed, and the adjustment order of each rule within the candidate rule adjustment section is rearranged according to the order of score change positions corresponding to rule stagnation sections. During the rearrangement process, rules closest to the score change positions of rule stagnation sections are prioritized, allowing the rule adjustment order to unfold gradually along the score change path in the score change sequence. In this way, a consistent correspondence is established between the rule adjustment order and the score change process in the score change sequence, enabling the rule adjustment process to progress gradually along the change path of the score change sequence.

[0044] After the rule adjustment order has been rearranged, a rule adjustment progression order is formed based on the rearranged rule order, and this progression order is used to limit the scope of subsequent rule content changes. Specifically, after the rule adjustment order has been rearranged according to the score change positions corresponding to the rule stagnation sections, the arrangement result can be organized into a rule adjustment progression order. In the rule adjustment progression order, each rule is arranged sequentially according to the score change path in the score change sequence, so that the rule content changes can gradually unfold along the score change direction presented by the score change sequence. Around the rule adjustment progression order, the corresponding positions in the score change sequence are organized segment by segment, so that the scope of rule content changes in the subsequent process can be limited to the score change position range corresponding to the rule adjustment progression order. The rule adjustment progression order formed in this way allows the rule content change process to advance gradually under the reference relationship formed by the score change sequence and the rule stagnation section, so that the scope of rule content changes always remains within the time range limited by the score change sequence, providing a continuous basis for subsequent reading of the subsequent change process of the score change sequence and adjusting the rule content segment by segment according to the rule adjustment progression order.

[0045] Based on the rule adjustment and advancement sequence, continue to read the subsequent changes in the score change sequence, adjust the rule content segment by segment in the time segment corresponding to the rule adjustment and advancement sequence, and keep the rule content unchanged after the score change returns to stability, so as to form a rule adaptive evolution process that is gradually advanced within the rigid constraint range; Based on the established sequence of rule adjustments, the process continues to focus on the subsequent changes in the scoring sequence. This ensures that rule changes proceed step-by-step within the timeframes defined by the rule adjustment sequence, and that the rule content remains unchanged once the scoring changes stabilize. This process, therefore, forms a gradual, adaptive evolutionary process within rigid constraints. The specific implementation steps are as follows: Based on the established rule adjustment sequence, the subsequent changes in the scoring change sequence are read according to this sequence, allowing the sequence to unfold continuously along the corresponding time path. Specifically, the rule adjustment sequence has been rearranged based on the scoring change positions corresponding to rule stagnation points, thus corresponding to a set of consecutive time segments within the scoring change sequence. Around these time segments, scoring records following them are read continuously, allowing the sequence to continue building upon the existing records. During this reading process, each scoring record in the sequence is connected chronologically, creating a continuous record of the scoring changes. This method ensures that subsequent changes in the scoring change sequence are fully presented within the time segments corresponding to the rule adjustment sequence, providing a basis for subsequent segmented rule adjustments.

[0046] Based on the continuous reading of subsequent changes in the scoring change sequence, the rule content changes are systematically organized segment by segment within the time intervals corresponding to the rule adjustment sequence, and the rule content is adjusted segment by segment according to the order defined by the rule adjustment sequence. Specifically, within the time intervals corresponding to the rule adjustment sequence, each time position corresponds to a scoring record and the corresponding rule content state in the scoring change sequence. Following the rule order arranged according to the rule adjustment sequence, the rule content at the corresponding position in the scoring change sequence is systematically organized, ensuring that the rule content changes unfold sequentially according to the rule adjustment sequence. During this process, each rule in the rule adjustment sequence is mapped to a corresponding time position in the scoring change sequence, and the rule content is adjusted segment by segment at the corresponding time position, allowing the rule content changes to unfold gradually along the scoring change path presented by the scoring change sequence. In this way, a continuous correspondence is established between the scoring change sequence and the rule adjustment sequence, ensuring that the rule content changes advance consistently in time.

[0047] After the rules have been adjusted segment by segment according to the rule adjustment sequence, the score change status in the score change sequence is continuously observed, and the rule content change process is further organized around the score change status in the score change sequence. When the score value in the score change sequence remains stable in multiple consecutive positions, this score change status can be regarded as a stage where the score change has returned to stability. In the stage where the score change has returned to stability, the rule content at the corresponding position in the score change sequence is kept in its original state, and no further rule content changes are made. During the organization process, the positions in the score change sequence that have maintained a stable state are uniformly marked, and the rule content status at the corresponding time position is recorded, so that the score change has returned to stability and the rule content has remained stable can be correlated. In this way, the stage where the score change has returned to stability can be clearly located in the score change sequence, thus providing a time reference for keeping the rule content unchanged.

[0048] Once a clear position has been established at the stage where score changes have stabilized, the state of the maintained rule content and the order of rule adjustments are unified and organized. This allows the rule content change process to form a progressive, adaptive evolutionary process within rigid constraints. Specifically, within the time frame defined by the rule adjustment order, rule content changes proceed segment by segment according to the order. After score changes stabilize, the rule content at the corresponding position in the score change sequence remains unchanged, preventing further changes. By unifying the rule adjustment order and the stable score change position, a continuous rule content change path is formed. The rule content change process starts from the initial stagnant segment, progresses through the candidate adjustment segment and the rule adjustment order, and finally stabilizes at the point where score changes have stabilized, maintaining the rule content unchanged. In this way, rule content changes are always within rigid constraints, and the rule content adjustment process progresses progressively with reference to the score change sequence, thus forming a progressive, adaptive evolutionary process within rigid constraints. This allows the rule system to continuously evolve and optimize its rule content while maintaining the stability of its core constraints.

[0049] This invention constructs a scoring change sequence and sets scoring stagnation references and rule stagnation sections within it, establishing a corresponding relationship between the scoring change process and the position where the rule applies. This allows for the identification of positions where the score continuously decreases while the rule content remains unchanged during the scoring change process. In this way, the scoring change trend in the scoring change sequence and the state of the rule content can be clearly mapped, so that the rule content change process no longer relies on experience-based judgment but is located and organized based on the scoring change process. This provides a clear reference for rule adjustment, reduces blind changes during rule adjustment, and allows rule content changes to unfold gradually around the scoring change state.

[0050] This invention generates candidate segments for rule adjustments within a score change sequence and further generates a rule adjustment progression order. This allows rule content changes to proceed segment by segment according to the time intervals defined by the score change sequence. Once the score changes stabilize, the rule content remains unchanged, thus keeping rule content changes within rigid constraints. In this way, the rule content change process forms a continuous progression path, enabling the rule adaptive evolution process to unfold gradually with reference to the score change sequence. This maintains the stability of the rule system during its evolution and allows the rule adjustment process to continuously optimize rule content within a defined progression range.

[0051] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A rule-adaptive evolutionary method based on rigid constraints, characterized in that, Includes the following steps: The output score records generated during the continuous operation of the intelligent agent are collected and organized into a score change sequence according to time order. The positions where the score repeatedly stops in the score change sequence are marked to form a score stop reference. By reviewing the time intervals corresponding to the scoring change sequence based on the scoring stagnation reference, the order in which each rule participated in the scoring within the time interval is sorted out segment by segment, and the positions where the score continues to decline while the rule content remains unchanged are identified and marked in the scoring change sequence to form rule stagnation segments; Based on the rule stagnation section, continue to expand the score change sequence, read the score change process before and after the rule stagnation section, sort out the score change rhythm within the rule stagnation section segment by segment, extract the positional relationship where the score has changed and the rule content remains unchanged, and form a rule adjustment candidate section in the score change sequence. By reviewing the time range corresponding to the score change sequence of the candidate segments for rule adjustment, the changes in rule content within the candidate segments are sorted out segment by segment, and the order of rule adjustment is rearranged according to the score change position of the rule stagnation segment to form the order of rule adjustment progress. Based on the rule adjustment and advancement sequence, continue reading the subsequent changes in the score change sequence, adjust the rule content segment by segment in the time interval corresponding to the rule adjustment and advancement sequence, and keep the rule content unchanged after the score change returns to stability.

2. The rule-adaptive evolutionary method based on rigid constraints according to claim 1, characterized in that, The steps for collecting data on the continuous operation of the intelligent agent, generating output score records, and forming a score change sequence are as follows: The system collects output score records generated during the continuous operation of the intelligent agent and arranges these records according to the order of score generation time to form a score change sequence. Based on the order of the score records in the score change sequence, the score value change status is sorted out segment by segment. The positions where the score value remains consistent in multiple consecutive positions are identified in the score change sequence and marked to form the positions where the score repeatedly stays. Based on the number of the repeated points of scoring in the scoring change sequence, the points of repeated scoring are centrally organized to form a scoring reference. Based on the reference position of the scoring stop, the time segments corresponding to the scoring change sequence are sorted out segment by segment, and the participation of the rule in the scoring in the corresponding time segment is recorded to form a correspondence between the position of the rule and the scoring change.

3. The rule-adaptive evolutionary method based on rigid constraints according to claim 2, characterized in that, The steps for reviewing the time intervals corresponding to the score change sequence around the score stagnation reference and forming regular stagnation intervals are as follows: The scoring change sequence is reviewed and organized around the corresponding position of the scoring stop reference in the scoring change sequence, and the scoring records before and after the scoring change sequence are continuously expanded according to the scoring stop reference position to form the time segment corresponding to the scoring stop reference. Based on the corresponding time intervals of the scoring reference, the participation of the corresponding rules in the scoring change sequence is sorted out segment by segment, and the order of participation of each rule in the scoring is arranged according to the time sequence of the scoring change sequence to form a record of the order of participation of the rules in the scoring. Based on the rules, the scoring order records are used to jointly organize the scoring change trend and rule content change status in the scoring change sequence, and the positional relationship of continuous score decline and unchanged rule content is identified in the scoring change sequence. Based on the positional relationship of continuously decreasing scores while the rule content remains unchanged, the consecutive positions corresponding to the score change sequence are marked and organized to form rule stagnation segments and present the correspondence between score stagnation and rule content.

4. The rule-adaptive evolutionary method based on rigid constraints according to claim 3, characterized in that, Based on the corresponding positions of the rule stagnation segments in the score change sequence, the positions in the score change sequence that continuously decrease in score are continuously organized, and the rule content states corresponding to the positions of continuous score decrease are recorded segment by segment to form a correspondence record between the continuous score decrease state and the rule content unchanged state.

5. The rule-adaptive evolutionary method based on rigid constraints according to claim 3, characterized in that, The steps for expanding the score change sequence based on the rule-stagnant segment and forming the rule adjustment candidate segment are as follows: The scoring change sequence is continuously expanded around the corresponding position of the rule stagnation section in the scoring change sequence, and the scoring records before and after the rule stagnation section are sequentially arranged to form a continuous scoring change section. Based on the continuous range of score changes, the score records within the rule stagnation range are organized segment by segment to form a record of the rhythm of score changes within the rule stagnation range. Based on the record of the rhythm of score changes within the rule stagnation zone, the score value change state and rule content state in the score change sequence are jointly organized, and the positional relationship of score changes and rule content unchanged is identified in the score change sequence. Based on the positional relationship where the scores have changed but the rule content remains unchanged, the corresponding consecutive positions in the score change sequence are sorted sequentially to form candidate segments for rule adjustment in the score change sequence.

6. The rule-adaptive evolutionary method based on rigid constraints according to claim 5, characterized in that, The steps for reviewing the score change sequence of candidate segments for rule adjustment and establishing the rule adjustment implementation order are as follows: The scoring change sequence is reviewed and organized around the corresponding position of the candidate segment for rule adjustment in the scoring change sequence. Then, the scoring records before and after the candidate segment for rule adjustment are continuously expanded according to the corresponding position of the candidate segment for rule adjustment to form the time range corresponding to the candidate segment for rule adjustment. Based on the rules, the time range corresponding to the candidate segments is adjusted, and the changes in the rule content corresponding to the score records in the score change sequence are sorted out segment by segment, and the rule content change records are formed according to the time order of the score change sequence. Based on the record of rule content changes and the corresponding score change positions in the rule stagnation section, the rule participation positions in the candidate rule adjustment section are organized accordingly, and the rule adjustment order is arranged according to the order of the score change positions in the rule stagnation section. The order of rule adjustment is determined by arranging the rules and then the corresponding positions of the score change sequence are sorted out segment by segment according to the rule adjustment order to limit the scope of rule content change.

7. The rule-adaptive evolutionary method based on rigid constraints according to claim 6, characterized in that, The corresponding positions of the scoring change sequence are continuously organized according to the rules to adjust the advancement order, and the scoring change status in the scoring change sequence is recorded in correspondence with the rule content change status. A rule content change path is formed in the scoring change sequence to limit the rule content change process to advance sequentially along the scoring change path.

8. The rule-adaptive evolutionary method based on rigid constraints according to claim 6, characterized in that, Based on the rules, the order of advancement is adjusted, and the subsequent changes in the score change sequence are read to form the rule-adaptive evolutionary process as follows: Based on the order of rule adjustments, the subsequent score records in the score change sequence are continuously read at the corresponding positions in the score change sequence to form the subsequent change process of the score change sequence; By combining the subsequent changes in the scoring change sequence with the corresponding time periods of the rule adjustment implementation order, the rule content at the corresponding position in the scoring change sequence is sorted out segment by segment, and the rule content is adjusted segment by segment according to the rule adjustment implementation order. The process of adjusting the rules segment by segment continuously organizes the status of score value changes in the score change sequence, and identifies the stable position of score change in the score change sequence. By combining the score change to restore the stable position, the rule content at the corresponding position in the score change sequence is maintained and organized, and the rule adaptive evolution process is gradually promoted within the rigid constraint range formed by the rule adjustment and advancement order.

Images