A method and system for optimizing urban planning based on multi-agent simulation

Abstract

This invention discloses a method and system for urban planning optimization based on multi-agent simulation, relating to the field of urban planning technology. By continuously tracking changes in the distribution structure of multi-agent behavior during simulation, if the discreteness sequence Div shows a sudden increase or continuous expansion trend at a certain time point, it is accurately located using the consistency collapse candidate point Cbr. Furthermore, the consistency collapse criterion Brk is used to determine whether the planning scheme has entered an uncontrollable state. Thus, the failure assessment result Evl is output before the actual implementation of the plan. This can guide planners to make targeted adjustments to the planning scheme without changing hardware conditions, avoiding real-world problems such as localized congestion, functional imbalance, or fragmented group behavior after the plan is implemented. This significantly improves the stability, controllability, and long-term adaptability of urban planning schemes in real-world operating environments.

Description

Technical Field

[0001] This invention relates to the field of urban planning technology, specifically to an urban planning optimization method and system based on multi-agent simulation. Background Technology

[0002] As cities continue to expand and their functional structures become increasingly complex, urban planning has gradually shifted from relying on empirical judgment and static indicator analysis to a decision-making approach based on computational models. Among these approaches, constructing simulation systems involving numerous stakeholders to dynamically model urban operations has become a crucial means of understanding urban complexity. Against this backdrop, as the simulation focus shifts from macroscopic structures to microscopic behaviors, accurately assessing the controllability of planning schemes in real-world environments has become a key challenge in the field.

[0003] In existing urban planning evaluation processes based on multi-agent simulations, the quality of planning schemes is typically measured by overall statistical indicators such as average travel time, average service accessibility, and average load level. These indicators can reflect the overall operational efficiency of the urban system to some extent, but they presuppose that individual behaviors maintain a relatively consistent response pattern after the plan is implemented. However, in actual simulations, due to differences in initial conditions, behavioral paths, and environmental perception, different agents often exhibit significant differentiation at certain stages, causing behavioral outcomes under the same planning conditions to become highly discrete rather than uniform. Existing evaluation methods, due to their over-reliance on average values, struggle to identify this critical state of transition from consistent to divergent behavior in a timely manner, thus failing to effectively determine whether a planning scheme is nearing failure. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for urban planning optimization based on multi-agent simulation, which solves the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a city planning optimization method based on multi-agent simulation, comprising the following steps:

[0006] S1. Collect behavioral data of multiple agents in an urban simulation environment, and construct a set of behavioral trajectories Beh representing the behavioral responses of planning schemes;

[0007] S2. Based on the set of behavioral trajectories Beh, construct a behavior distribution table Dst that reflects the behavior distribution structure of multiple agents;

[0008] S3. Based on the behavior distribution table Dst, calculate the discreteness change of the multi-agent behavior distribution to form a discreteness sequence Div, and extract the candidate time positions where behavioral consistency changes abruptly in the discreteness sequence Div to form a consistency breakdown candidate point Cbr.

[0009] S4. Based on the discrete sequence Div, construct the consistency breakdown criterion Brk, determine whether the consistency breakdown candidate point Cbr satisfies the behavioral consistency breakdown condition, and generate the failure evaluation result Evl to represent the controllable state of the planning scheme.

[0010] S5. Based on the failure assessment result Evl, perform simulation optimization on the urban planning scheme and output the planning optimization output Opt to represent the optimization result.

[0011] Preferably, S1 includes S11;

[0012] S11. In an urban simulation environment, the behavior of various intelligent agents participating in the simulation is monitored, and the original behavior records generated by the multiple intelligent agents during the simulation operation are collected.

[0013] The original behavior record includes information on the agent's spatial location change, the time of the behavior, the duration of the behavior, and the behavior type identifier.

[0014] The collected raw behavior records are standardized according to a unified time reference and spatial coordinate system, and merged according to the agent identifier to form a behavior record set Rec representing the original behavior state of multiple agents.

[0015] Preferably, S1 further includes S12;

[0016] S12. Based on the behavior record set Rec, according to the preset time continuity rules and spatial coherence rules, the continuous behavior records of the same agent are associated and reconstructed to generate a behavior trajectory that reflects the continuous behavior evolution process of the agent within the simulation cycle.

[0017] The reconstructed behavioral trajectories of each agent are organized and summarized according to time windows and spatial units to form a set of behavioral trajectories Beh, which represents the behavioral response characteristics of multiple agents under the planning scheme.

[0018] Preferably, S2 includes S21;

[0019] S21. Based on the set of behavioral trajectories Beh, the behavioral trajectories of multiple agents are grouped and statistically analyzed according to a preset statistical time window and spatial analysis unit. The behavioral results of multiple agents within the same time window and the same spatial analysis unit are aggregated to form a distribution description for representing the structure of group behavior.

[0020] The behavior distribution table Dst includes the following distribution contents: the distribution structure of multi-agent behavior trajectories in the activity duration dimension, the distribution structure of multi-agent behavior trajectories in the path selection feature dimension, and the distribution structure of multi-agent behavior trajectories in the activity chain integrity dimension.

[0021] Preferably, S3 includes S31;

[0022] S31. Based on the behavior distribution table Dst, and according to the time sequence of the statistical time windows, calculate the dispersion of the following distribution contents for the multi-agent behavior distribution structure within each statistical time window:

[0023] The distribution structure of multi-agent behavioral trajectories along the activity duration dimension;

[0024] The distribution structure of multi-agent behavioral trajectories along the path selection feature dimension;

[0025] The distribution structure of multi-agent behavioral trajectories along the activity chain integrity dimension;

[0026] The discreteness calculation for each type of distribution structure is performed using a unified distribution consistency evaluation method. The distribution consistency evaluation method includes: within the same statistical time window, based on the overall dispersion of the behavioral values ​​of each agent in the corresponding distribution structure, evaluating the degree of deviation of the distribution structure from the centralized state, and obtaining the single discreteness result corresponding to the distribution structure.

[0027] Subsequently, the dispersion results corresponding to the three types of distribution structures within the same statistical time window are synthesized to obtain a comprehensive dispersion result that represents the overall degree of differentiation of multi-agent behavior within the statistical time window.

[0028] The comprehensive dispersion results corresponding to each statistical time window are arranged in chronological order to form a dispersion sequence Div that describes the evolution of the consistency of multi-agent behavior over time.

[0029] Preferably, S3 further includes S32;

[0030] S32. Based on the discreteness sequence Div, the discreteness results corresponding to adjacent statistical time windows are compared one by one according to the time order of the statistical time windows to obtain the continuous change relationship of discreteness with time.

[0031] In the discrete sequence Div, when the discrete result corresponding to a certain statistical time window suddenly increases compared to the discrete result corresponding to the previous statistical time window, and the sudden increase continues to exist in at least one subsequent adjacent statistical time window, the statistical time window is marked as a candidate time position where the behavioral consistency changes abruptly.

[0032] Meanwhile, in the discrete sequence Div, when the discrete result shows a unidirectional expansion trend in multiple consecutive statistical time windows, that is, when the discrete result corresponding to the subsequent statistical time window is not lower than the discrete result corresponding to the previous statistical time window, the starting statistical time window of the continuous expansion trend is marked as the candidate time position where the behavioral consistency changes abruptly.

[0033] The candidate time locations will be aggregated to form a consistency breakdown candidate point Cbr, which represents the potential risk of consistency breakdown in the behavior of multiple agents.

[0034] The sudden increase state refers to an abnormal jump in the discrete result corresponding to a certain statistical time window, which exceeds a preset proportion range relative to the normal change range of the discrete sequence Div in the stable phase.

[0035] The unidirectional expansion trend refers to the evolutionary state in which the dispersion result shows a continuous expansion without any pullback within multiple consecutive statistical time windows.

[0036] Preferably, S4 includes S41;

[0037] S41. Based on the discrete sequence Div, select a continuous statistical time window located in the stage where the behavioral consistency is in a stable state as the benchmark analysis segment.

[0038] Within the benchmark analysis section, the level of change in the statistical dispersion results is used to determine the dispersion reference range corresponding to the stable phase.

[0039] Based on the aforementioned dispersion reference range, a consistency breakdown criterion Brk is constructed to determine whether behavioral consistency has substantially broken down.

[0040] The consistency breakdown criterion Brk is used to determine that when the discrete sequence Div continuously deviates from the discrete reference range and remains in an expanding state at the time position corresponding to the consistency breakdown candidate point Cbr, the behavioral consistency is determined to enter a breakdown state.

[0041] Preferably, S4 further includes S42;

[0042] S42. For the consistency breakdown candidate point Cbr, call the consistency breakdown criterion Brk to determine whether the corresponding statistical time window meets the behavior consistency breakdown condition.

[0043] When the consistency breakdown criterion Brk is met, the corresponding planning scheme is determined to enter an uncontrollable state within the statistical time window and is marked as a failure state in the failure evaluation result Evl; when the consistency breakdown criterion Brk is not met, the corresponding planning scheme is determined to remain in a controllable state within the statistical time window and is marked as a controllable state in the failure evaluation result Evl.

[0044] Preferably, S5 includes S51;

[0045] S51. Based on the failure assessment result Evl, analyze the operation status of the urban planning scheme during the simulation process, and identify the time position that leads the planning scheme into an uncontrollable state and the corresponding behavioral consistency collapse characteristics.

[0046] For the planning schemes marked as uncontrollable in the failure assessment result Evl, perform structural adjustments at the simulation level, and re-execute steps S1 to S4 to obtain the adjusted failure assessment result Evl.

[0047] The failure assessment result Evl corresponding to the adjusted planning scheme is compared with the failure assessment result Evl of the original planning scheme. The planning scheme in which the behavioral consistency collapse is eliminated is selected as the optimization result, and the planning optimization output Opt is output to represent the optimization effect of the planning scheme.

[0048] The planning optimization output Opt includes the controllable state determination result of the planning scheme, the change of the behavior consistency collapse point, and the corresponding discrete sequence Div.

[0049] A city planning optimization system based on multi-agent simulation includes a simulation data acquisition module, a distributed extraction module, a distributed discrete computing module, a consistency judgment module, and an optimization decision module.

[0050] The simulation data acquisition module collects behavioral data of multiple agents in an urban simulation environment and constructs a set of behavioral trajectories Beh representing the behavioral responses of planning schemes.

[0051] The distribution extraction module constructs a behavior distribution table Dst that reflects the behavior distribution structure of multiple agents based on the behavior trajectory set Beh;

[0052] The distributed discrete computing module calculates the discreteness change of the multi-agent behavior distribution according to the behavior distribution table Dst, forming a discreteness sequence Div, and extracts the candidate time positions where behavioral consistency changes abruptly in the discreteness sequence Div, forming a consistency collapse candidate point Cbr.

[0053] The consistency judgment module constructs a consistency breakdown criterion Brk based on the discrete sequence Div, determines whether the consistency breakdown candidate point Cbr satisfies the behavioral consistency breakdown condition, and generates a failure assessment result Evl to represent the controllable state of the planning scheme.

[0054] The optimization decision module performs simulation optimization of the urban planning scheme based on the failure assessment result Evl, and outputs the planning optimization output Opt to represent the optimization result.

[0055] This invention provides a method and system for urban planning optimization based on multi-agent simulation, which has the following beneficial effects:

[0056] (1) By continuously tracking the changes in the distribution structure of multi-agent behavior during the simulation process, once the discrete sequence Div shows a sudden increase or continuous expansion trend at a certain time position, it is accurately located by the consensus collapse candidate point Cbr, and further judged by the consensus collapse criterion Brk whether the planning scheme has entered an uncontrollable state, so as to output the failure assessment result Evl before the planning is actually implemented. This guides planners to make targeted adjustments to the planning scheme without changing the hardware conditions, so as to avoid real problems such as local congestion and loss of control, functional area imbalance or group behavior split after the planning is implemented.

[0057] (2) By further structuring the behavioral trajectory set Beh into a behavioral distribution table Dst containing multi-dimensional behavioral features, and on this basis constructing a continuous and comparable discrete sequence Div and a consistency breakdown candidate point Cbr under explicit judgment rules, this method achieves a refined identification capability "from structure to evolution" in urban planning simulation evaluation. Its outstanding technical effect lies in its ability to capture the process signal of the evolution of group behavior from stability to disorder in advance, rather than only discovering problems at the result level. Unlike the existing technology that only performs static evaluation on a certain moment or a certain indicator, this scheme, through the collaborative analysis of the activity duration distribution structure, path selection feature distribution structure, and activity chain integrity distribution structure, enables the discrete sequence Div to truly reflect the evolution trajectory of multi-agent behavioral consistency over time, significantly improving the sensitivity and foresight of urban planning simulation evaluation to complex changes in reality.

[0058] (3) By selecting a benchmark analysis segment in the discrete sequence Div where behavioral consistency is stable, a consistency breakdown criterion Brk, which is adaptive to the specific planning scenario, is constructed. This method avoids the risk of misjudgment caused by using fixed thresholds or empirical standards, ensuring that the determination of behavioral consistency breakdown is always based on the normal operation characteristics of the planning scheme itself. When the consistency breakdown candidate point Cbr is confirmed to be valid under the consistency breakdown criterion Brk, the failure assessment result Evl not only gives the conclusion that the planning scheme has entered an uncontrollable state, but also clearly marks the time location range of the behavioral consistency breakdown and its deviation characteristics relative to the discrete reference range, thus providing accurate guidance for subsequent optimization. In real urban planning simulation, it significantly improves the stability, controllability, and decision credibility of urban planning schemes in complex real operating environments. Attached Figure Description

[0059] Figure 1 This is a schematic diagram illustrating the steps of an urban planning optimization method based on multi-agent simulation according to the present invention.

[0060] Figure 2 This is a schematic diagram of a city planning optimization system based on multi-agent simulation according to the present invention. Detailed Implementation

[0061] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0062] Example 1

[0063] This invention provides a method for urban planning optimization based on multi-agent simulation. Please refer to [link / reference]. Figure 1 This includes the following steps:

[0064] S1. Collect behavioral data of multiple agents in an urban simulation environment, and construct a set of behavioral trajectories Beh representing the behavioral responses of planning schemes;

[0065] S2. Based on the set of behavioral trajectories Beh, construct a behavior distribution table Dst that reflects the behavior distribution structure of multiple agents;

[0066] S3. Based on the behavior distribution table Dst, calculate the discreteness change of the multi-agent behavior distribution to form a discreteness sequence Div, and extract the candidate time positions where behavioral consistency changes abruptly in the discreteness sequence Div to form a consistency breakdown candidate point Cbr.

[0067] S4. Based on the discrete sequence Div, construct the consistency breakdown criterion Brk, determine whether the consistency breakdown candidate point Cbr satisfies the behavioral consistency breakdown condition, and generate the failure evaluation result Evl to represent the controllable state of the planning scheme.

[0068] S5. Based on the failure assessment result Evl, perform simulation optimization on the urban planning scheme and output the planning optimization output Opt to represent the optimization result.

[0069] In this embodiment, a multi-agent simulation-based urban planning optimization method, consisting of steps S1 to S5, introduces a set of behavioral trajectories (Beh), a behavioral distribution table (Dst), a discrete sequence (Div), a consensus collapse candidate point (Cbr), a consensus collapse criterion (Brk), a failure assessment result (Evl), and a planning optimization output (Opt). This method achieves a substantial improvement in urban planning schemes from "overall average effectiveness" to "controllable group behavior." Unlike existing technologies that rely solely on macroscopic indicators such as average travel time and average load levels, this method continuously tracks changes in the multi-agent behavioral distribution structure during simulation. Once the discrete sequence (Div) shows a sudden increase or continuous expansion trend at a certain time point, it is precisely located using the consensus collapse candidate point (Cbr). Furthermore, the consensus collapse criterion (Brk) is used to determine whether the planning scheme has entered an uncontrollable state, thus outputting the failure assessment result (Evl) before the planning is actually implemented. Taking real-world urban planning as an example, in the simulation phase of functional layout adjustment or traffic organization optimization in new areas, traditional methods may show good overall operational efficiency. However, this method can identify early abnormal behaviors such as frequent interruptions in the activity chains of some residents' intelligent agents and highly differentiated path choices. This abnormality is reflected as a significant change in the discreteness sequence Div, and ultimately, the failure assessment result Evl clearly indicates that the planning scheme has a potential risk of getting out of control. Based on this, the planning optimization output Opt generated in step S5 can guide planners to make targeted adjustments to the planning scheme without changing the hardware conditions. This avoids real-world problems such as local congestion, functional imbalance, or fragmented group behavior after the plan is implemented, thereby significantly improving the stability, controllability, and long-term adaptability of urban planning schemes in real operating environments. This directly solves the technical shortcomings mentioned in the background, such as "average indicators masking the risk of group differentiation and the difficulty in timely detection of planning failures."

[0070] Example 2

[0071] Specifically: S1 includes S11;

[0072] S11. In an urban simulation environment, the behavior of various intelligent agents participating in the simulation is monitored, and the original behavior records generated by the multiple intelligent agents during the simulation operation are collected.

[0073] The original behavior record includes information on the agent's spatial location change, the time of the behavior, the duration of the behavior, and the behavior type identifier.

[0074] The collected raw behavior records are standardized according to a unified time reference and spatial coordinate system, and merged according to the agent identifier to form a behavior record set Rec representing the original behavior state of multiple agents;

[0075] Among them, the behavior record set Rec serves as the basic input data for subsequent behavior trajectory construction.

[0076] S1 further includes S12;

[0077] S12. Based on the behavior record set Rec, according to the preset time continuity rules and spatial coherence rules, the continuous behavior records of the same agent are associated and reconstructed to generate a behavior trajectory that reflects the continuous behavior evolution process of the agent within the simulation cycle.

[0078] The reconstructed behavioral trajectories of each agent are organized and summarized according to time windows and spatial units to form a set of behavioral trajectories Beh, which represents the behavioral response characteristics of multiple agents under the planning scheme.

[0079] The behavioral trajectory set Beh serves as the direct input object for subsequent behavioral distribution construction and consistency analysis.

[0080] It should be noted that:

[0081] The preset temporal continuity rule is used to determine whether adjacent behavior records of the same agent belong to the same continuous behavior trajectory. Its specific implementation includes:

[0082] Sort the behavior records belonging to the same agent in the behavior record set Rec by time of occurrence.

[0083] For two adjacent behavior records after time sorting, determine whether the time interval is less than or equal to a preset continuous time threshold;

[0084] When the time interval between two adjacent behavior records meets the continuous time threshold condition, it is determined that the two behavior records meet the time continuity requirement and are allowed to proceed to subsequent association processing;

[0085] When the time interval between two adjacent behavior records exceeds the continuous time threshold, the time continuity is determined to be interrupted, and the subsequent behavior record is used as the starting point of the new behavior trajectory.

[0086] The continuous time threshold is set before the simulation starts based on the simulation time step and the agent behavior update frequency, and remains consistent throughout the entire simulation cycle.

[0087] The preset spatial coherence rules are used to limit whether temporally continuous behavior records have a reasonable movement or activity continuity relationship in the spatial dimension. Specific implementation methods include:

[0088] Based on the spatial location information recorded in the behavior record set Rec, calculate the spatial location change distance between adjacent behavior records;

[0089] Determine whether the distance of the spatial position change is less than or equal to a preset spatial coherence threshold;

[0090] When the distance of the change in spatial location meets the spatial coherence threshold condition, it is determined that the behavior record has spatial coherence.

[0091] When the distance of the change in spatial location exceeds the spatial continuity threshold, it is determined that the behavior record does not have spatial continuity, and the continuation of the current trajectory is terminated.

[0092] The spatial coherence threshold is set based on the spatial unit scale and the maximum reasonable movement speed of the agent in the urban simulation environment.

[0093] Specifically, the association involved in associating and reconstructing continuous behavior records of the same intelligent agent includes:

[0094] Only pairs of behavior records that simultaneously satisfy the temporal continuity rule and the spatial coherence rule are associated;

[0095] The behavior records that meet the association conditions are connected sequentially in chronological order to form an initial chain of associated behaviors;

[0096] For behavior records that do not meet the association conditions, terminate the current association behavior chain and mark it as the starting record of a new behavior trajectory;

[0097] The reconstruction process in associating and reconstructing continuous behavior records of the same agent specifically includes:

[0098] Using each initial chain of related behaviors as the basic unit, the behavior records within the chain are sequentially organized to generate continuous behavior trajectory segments.

[0099] Short time-space gaps in the behavior trajectory segments are filled in to maintain the integrity of the behavior trajectory in the time dimension.

[0100] The reconstructed behavior trajectory segments are subjected to consistency verification to ensure that there are no conflicts in the temporal order and spatial change direction of the behavior trajectory segments.

[0101] The verified behavior trajectory segments are confirmed as valid behavior trajectories of a single intelligent agent.

[0102] In this embodiment, by standardizing the collection of original behavior records of multiple agents, imposing strict temporal continuity and spatial coherence rules, and combining a clear association and reconstruction processing mechanism, the resulting behavior trajectory set Beh can realistically reflect the continuous behavioral evolution of agents within the urban simulation cycle. Its outstanding technical effect lies in significantly reducing the risk of behavioral distortion caused by data fragmentation, behavioral jumps, or simulation noise. Unlike existing methods that directly rely on discrete time-point locations or simple trajectory splicing, this solution uses the behavior record set Rec as an intermediate foundation layer to perform temporal sorting, threshold determination, and spatial rationality verification of the same agent's behavior, avoiding misjudging discontinuous or unreasonable behaviors as continuous behavioral trajectories. For example, in real-world urban simulations, resident agents may experience sudden position jumps due to simulation step size, path updates, or short-term pauses. Traditional methods easily misclassify such jumps as belonging to the same travel trajectory, while this method, through dual constraints of continuous temporal thresholds and spatial coherence thresholds, can accurately identify and cut off unreasonable trajectory segments, retaining only behavioral trajectory fragments that conform to real movement logic. The resulting set of behavioral trajectories, Beh, can effectively avoid problems such as abnormal activity duration, distorted path selection, or false interruption of activity chains caused by trajectory splicing errors when constructing the behavior distribution table Dst. This provides stable and reliable behavioral foundation data for the entire multi-agent simulation analysis. This advantage is particularly evident in urban planning scenarios with complex resident travel patterns, coarse simulation step size, or long simulation cycle.

[0103] Example 3

[0104] Specifically: S2 includes S21;

[0105] S21. Based on the set of behavioral trajectories Beh, the behavioral trajectories of multiple agents are grouped and statistically analyzed according to a preset statistical time window and spatial analysis unit. The behavioral results of multiple agents within the same time window and the same spatial analysis unit are aggregated to form a distribution description for representing the structure of group behavior.

[0106] The behavior distribution table Dst includes the following distribution contents: the distribution structure of multi-agent behavior trajectories in the activity duration dimension, the distribution structure of multi-agent behavior trajectories in the path selection feature dimension, and the distribution structure of multi-agent behavior trajectories in the activity chain integrity dimension;

[0107] By constructing the distribution as described above, the behavior distribution table Dst reflects the behavioral response structure of multiple agents under the planning scheme in the form of a distribution pattern rather than a single statistical value, providing a basic input for subsequent analysis based on the degree of distribution dispersion.

[0108] It should be noted that the activity duration distribution structure refers to the value distribution pattern formed based on the actual duration of the activities of each agent in the behavioral trajectory set Beh within the same statistical time window and the same spatial analysis unit.

[0109] Specifically, the activity duration distribution structure is obtained in the following way:

[0110] Extract the start and end times of the activity for each behavior trajectory from the set of behavior trajectories Beh;

[0111] The corresponding activity duration is calculated.

[0112] The duration of activities within the same statistical time window and the same spatial analysis unit is aggregated to form a distribution structure that reflects the concentration, dispersion, and extreme value distribution of the duration of activities.

[0113] The activity duration distribution structure is used to characterize whether the activity completion rhythm of multiple agents tends to be consistent or begins to show obvious differentiation under the influence of planning schemes.

[0114] The path selection feature distribution structure refers to the distribution pattern formed based on the spatial movement characteristics of the behavioral trajectories of each agent in the behavioral trajectory set Beh within the same statistical time window and the same spatial analysis unit.

[0115] Specifically, the path selection feature distribution structure is obtained in the following way:

[0116] Extract the path sequence corresponding to each behavior trajectory from the behavior trajectory set Beh;

[0117] Path selection features such as path length, path detour degree, and number of path turns are identified based on path sequence;

[0118] By aggregating path selection characteristics within the same statistical time window and the same spatial analysis unit, a distribution structure reflecting the convergence or dispersion of path selection is formed;

[0119] The path selection feature distribution structure is used to characterize whether multiple agents still exhibit similar path decisions or begin to show significantly differentiated path selection behaviors under the same planning conditions.

[0120] The activity chain integrity distribution structure refers to the distribution pattern formed based on the completion status of each agent's activity chain in the behavioral trajectory set Beh within the same statistical time window and the same spatial analysis unit.

[0121] Specifically, the activity chain integrity distribution structure is obtained in the following way:

[0122] Identify the activity chain structure corresponding to each behavior trajectory from the behavior trajectory set Beh;

[0123] Determine if there are any instances of activity interruption, activity skipping, or forced termination in the activity chain;

[0124] The integrity status of activity chains within the same statistical time window and the same spatial analysis unit is aggregated to form a distribution structure that reflects the differences in the distribution of activity chain integrity.

[0125] The activity chain integrity distribution structure is used to characterize whether the activity organization of multiple agents remains stable as a whole under the action of the planning scheme, or whether there is a trend of local group activity chain failure.

[0126] Among them, aggregation processing refers to the process of uniformly collecting and organizing multi-agent behavioral data that meet the same statistical time window and the same spatial analysis unit conditions during the construction of the behavior distribution table Dst.

[0127] Specifically, the polymerization process includes:

[0128] Multiple behavioral trajectories from the behavioral trajectory set Beh are classified and grouped according to statistical time windows and spatial analysis units;

[0129] Within each group, only the behavioral feature values ​​required to construct the distribution structure are retained, without averaging or single-value compression.

[0130] The distribution pattern is used as the output of the aggregation result for subsequent behavior dispersion analysis;

[0131] Through the aggregation process, the behavior distribution table Dst is ensured to retain the structural information of group behavior differences, rather than being compressed into an overall statistical indicator that cannot reflect the differentiation trend.

[0132] S3 includes S31;

[0133] S31. Based on the behavior distribution table Dst, and according to the time sequence of the statistical time windows, calculate the dispersion of the following distribution contents for the multi-agent behavior distribution structure within each statistical time window:

[0134] The distribution structure of multi-agent behavioral trajectories along the activity duration dimension;

[0135] The distribution structure of multi-agent behavioral trajectories along the path selection feature dimension;

[0136] The distribution structure of multi-agent behavioral trajectories along the activity chain integrity dimension;

[0137] The discreteness calculation for each type of distribution structure is performed using a unified distribution consistency evaluation method. The distribution consistency evaluation method includes: within the same statistical time window, based on the overall dispersion of the behavioral values ​​of each agent in the corresponding distribution structure, evaluating the degree of deviation of the distribution structure from the centralized state, and obtaining the single discreteness result corresponding to the distribution structure.

[0138] Subsequently, the dispersion results corresponding to the three types of distribution structures within the same statistical time window are synthesized to obtain a comprehensive dispersion result that represents the overall degree of differentiation of multi-agent behavior within the statistical time window.

[0139] The comprehensive dispersion results corresponding to each statistical time window are arranged in chronological order to form a dispersion sequence Div that describes the evolution of the consistency of multi-agent behavior over time.

[0140] S3 further includes S32;

[0141] S32. Based on the discreteness sequence Div, the discreteness results corresponding to adjacent statistical time windows are compared one by one according to the time order of the statistical time windows to obtain the continuous change relationship of discreteness with time.

[0142] In the discrete sequence Div, when the discrete result corresponding to a certain statistical time window suddenly increases compared to the discrete result corresponding to the previous statistical time window, and the sudden increase continues to exist in at least one subsequent adjacent statistical time window, the statistical time window is marked as a candidate time position where the behavioral consistency changes abruptly.

[0143] Meanwhile, in the discrete sequence Div, when the discrete result shows a unidirectional expansion trend in multiple consecutive statistical time windows, that is, when the discrete result corresponding to the subsequent statistical time window is not lower than the discrete result corresponding to the previous statistical time window, the starting statistical time window of the continuous expansion trend is marked as the candidate time position where the behavioral consistency changes abruptly.

[0144] The candidate time locations will be aggregated to form a consistency breakdown candidate point Cbr, which represents the potential risk of consistency breakdown in the behavior of multiple agents.

[0145] The sudden increase refers to an abnormal jump in the dispersion result corresponding to a certain statistical time window, which exceeds a preset proportion range relative to the normal change range of the dispersion sequence Div during the stable phase. The determination method is limited as follows:

[0146] First, a continuous statistical time window in a relatively stable state is selected from the discrete sequence Div as a stable phase to represent the natural fluctuation level of the discrete results under non-abnormal conditions.

[0147] During the stable phase, the variation range of the dispersion results between adjacent statistical time windows is statistically analyzed to obtain the normal variation range corresponding to the stable phase.

[0148] The dispersion result corresponding to the current statistical time window is compared with the dispersion result corresponding to the previous statistical time window. When the change amplitude exceeds more than twice the normal change amplitude, it is determined that the dispersion result corresponding to the current statistical time window is in a state of sudden increase.

[0149] In the discrete sequence Div, the continued existence of the sudden increase state means that after a sudden increase state is determined to occur within a certain statistical time window:

[0150] Within at least one immediately following statistical time window, the corresponding dispersion result did not fall back to within the normal range of variation during the stable phase;

[0151] Or, within a continuous subsequent statistical time window, the dispersion result remains within an range no less than the dispersion result corresponding to the sudden increase state minus the normal change amplitude;

[0152] When any of the above conditions are met, the sudden increase is considered to be persistent;

[0153] The unidirectional expansion trend refers to the evolutionary state in which the dispersion result shows a continuous expansion without reversal within multiple consecutive statistical time windows. The determination method is limited as follows:

[0154] The dispersion results corresponding to adjacent statistical time windows in the dispersion sequence Div are compared according to the statistical time window order.

[0155] When the dispersion result of subsequent statistical time windows is not less than half of the dispersion result of the previous statistical time window minus the normal change range of the stable phase within at least two consecutive statistical time windows, the dispersion result is judged to be in a unidirectional expansion trend.

[0156] The starting statistical time window of a continuous segment that meets the above conditions is identified as the starting position of the unidirectional expansion trend.

[0157] By imposing the above restrictions, minor normal fluctuations are allowed in the unidirectional expansion trend, but obvious retracements are excluded, thus strictly distinguishing it from random noise.

[0158] In this embodiment, by further structuring the behavioral trajectory set Beh into a behavioral distribution table Dst containing multi-dimensional behavioral features, and on this basis constructing a continuously comparable discrete sequence Div and a consistency breakdown candidate point Cbr under explicit judgment rules, this method achieves a refined identification capability "from structure to evolution" in urban planning simulation evaluation. Its outstanding technical effect lies in its ability to capture the process signals of the evolution of group behavior from stability to disorder in advance, rather than only discovering problems at the result level. Unlike existing technologies that only perform static evaluation on a certain moment or a certain indicator, this scheme, through the collaborative analysis of the activity duration distribution structure, path selection feature distribution structure, and activity chain integrity distribution structure, enables the discrete sequence Div to truly reflect the evolution trajectory of multi-agent behavioral consistency over time. For example, when simulating traffic organization schemes for a certain area in real-world urban planning, traditional methods may only show that overall traffic efficiency remains stable. However, this method can detect, through the behavior distribution table Dst, that the distribution of path selection characteristics begins to widen significantly during certain time periods, while the proportion of interruptions in the activity chain integrity distribution structure increases. This is further manifested in the discreteness sequence Div as a sudden increase or unidirectional expansion trend within a continuous statistical time window, thus marking potential risk moments in advance in the consistency collapse candidate point Cbr. This identification method based on distribution structure and evolution trend allows planners to identify the critical stage of "group differentiation but not yet collapse" before the scheme reaches full failure. This provides a reliable basis for subsequent accurate judgment and adjustment through the consistency collapse criterion Brk and failure assessment results Evl, significantly improving the sensitivity and foresight of urban planning simulation assessment to complex changes in real-world behavior.

[0159] Example 4

[0160] Specifically: S4 includes S41;

[0161] S41. Based on the discrete sequence Div, select a continuous statistical time window located in the stage where the behavioral consistency is in a stable state as the benchmark analysis segment to reflect the discrete evolution characteristics of the multi-agent behavior distribution under normal operating conditions.

[0162] Within the benchmark analysis section, the level of change in the statistical dispersion results is used to determine the dispersion reference range corresponding to the stable phase.

[0163] Based on the aforementioned dispersion reference range, a consistency breakdown criterion Brk is constructed to determine whether behavioral consistency has substantially broken down.

[0164] The consistency breakdown criterion Brk is used to determine that when the discrete sequence Div continuously deviates from the discrete reference range and remains in an expanding state at the time position corresponding to the consistency breakdown candidate point Cbr, the behavioral consistency is determined to enter a breakdown state.

[0165] S4 also includes S42;

[0166] S42. For the consistency breakdown candidate point Cbr, call the consistency breakdown criterion Brk to determine whether the corresponding statistical time window meets the behavior consistency breakdown condition.

[0167] When the consistency breakdown criterion Brk is met, the corresponding planning scheme is determined to enter an uncontrollable state in the statistical time window and is marked as a failure state in the failure evaluation result Evl; when the consistency breakdown criterion Brk is not met, the corresponding planning scheme is determined to remain in a controllable state in the statistical time window and is marked as a controllable state in the failure evaluation result Evl.

[0168] The failure assessment result Evl includes the controllable state determination result corresponding to the planning scheme, the time location range of the behavior consistency collapse, and the deviation feature description of the discreteness sequence Div relative to the discreteness reference range.

[0169] S5 includes S51;

[0170] S51. Based on the failure assessment result Evl, analyze the operation status of the urban planning scheme during the simulation process, and identify the time position that leads the planning scheme into an uncontrollable state and the corresponding behavioral consistency collapse characteristics.

[0171] For the planning schemes marked as uncontrollable in the failure assessment result Evl, perform structural adjustments at the simulation level, and re-execute steps S1 to S4 to obtain the adjusted failure assessment result Evl.

[0172] The failure assessment result Evl corresponding to the adjusted planning scheme is compared with the failure assessment result Evl of the original planning scheme. The planning scheme in which the behavioral consistency collapse is eliminated is selected as the optimization result, and the planning optimization output Opt is output to represent the optimization effect of the planning scheme.

[0173] The planning optimization output Opt includes the controllable state determination result of the planning scheme, the change of the behavior consistency collapse point, and the corresponding discrete sequence Div.

[0174] In this embodiment, through the back-end judgment and optimization process consisting of steps S41, S42, and S51, this method achieves a closed-loop decision-making capability from "risk identification" to "controllability repair" in urban planning simulation. Its significant technical effect lies in its ability to directly transform changes in the consistency of multi-agent behavior into a clear judgment of whether a planning scheme is controllable, further guiding targeted optimization of the planning scheme. By selecting a benchmark analysis segment in the discrete sequence Div where behavioral consistency is stable, and constructing a consistency collapse criterion Brk that is adaptive to the specific planning scenario, this method avoids the risk of misjudgment caused by using fixed thresholds or empirical standards, ensuring that the judgment of behavioral consistency collapse is always based on the normal operating characteristics of the planning scheme itself. When the consistency collapse candidate point Cbr is confirmed under the consistency collapse criterion Brk, the failure assessment result Evl not only gives the conclusion that the planning scheme has entered an uncontrollable state, but also clearly marks the time range of the behavioral consistency collapse and its deviation characteristics relative to the discrete reference range, thus providing precise guidance for subsequent optimization. In real-world urban planning simulations, such as when comparing multiple schemes for the functional layout or traffic organization of newly developed areas, traditional methods often only allow for passive adjustments after the overall indicators of the scheme deteriorate. This method, however, can directly pinpoint the critical period leading to uncontrolled group behavior based on the failure assessment result (Evl). Step S51 performs structural adjustments to the planning scheme at the simulation level, and steps S1 to S4 are executed again for verification. The final output, the optimized planning output Opt, not only indicates whether the planning scheme has returned to a controllable state but also intuitively reflects the changes in the behavioral consistency breakdown point and the improvement trend of the dispersion sequence Div. This allows planners to complete multiple rounds of self-verifying optimization before implementation, significantly improving the stability, controllability, and decision-making credibility of urban planning schemes in complex real-world operating environments.

[0175] Example 5

[0176] A city planning optimization system based on multi-agent simulation, please refer to... Figure 2 Specifically, it includes a simulation data acquisition module, a distributed extraction module, a distributed discrete calculation module, a consistency judgment module, and an optimization decision module;

[0177] The simulation data acquisition module collects behavioral data of multiple agents in an urban simulation environment and constructs a set of behavioral trajectories Beh representing the behavioral responses of planning schemes.

[0178] The distribution extraction module constructs a behavior distribution table Dst that reflects the behavior distribution structure of multiple agents based on the behavior trajectory set Beh;

[0179] The distributed discrete computing module calculates the discreteness change of the multi-agent behavior distribution according to the behavior distribution table Dst, forming a discreteness sequence Div, and extracts the candidate time positions where behavioral consistency changes abruptly in the discreteness sequence Div, forming a consistency collapse candidate point Cbr.

[0180] The consistency judgment module constructs a consistency breakdown criterion Brk based on the discrete sequence Div, determines whether the consistency breakdown candidate point Cbr satisfies the behavioral consistency breakdown condition, and generates a failure assessment result Evl to represent the controllable state of the planning scheme.

[0181] The optimization decision module performs simulation optimization of the urban planning scheme based on the failure assessment result Evl, and outputs the planning optimization output Opt to represent the optimization result.

[0182] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing urban planning based on multi-agent simulation, characterized in that: Includes the following steps: S1. Collect behavioral data of multiple agents in an urban simulation environment, and construct a set of behavioral trajectories Beh representing the behavioral responses of planning schemes; S2. Based on the set of behavioral trajectories Beh, construct a behavior distribution table Dst that reflects the behavior distribution structure of multiple agents; S3. Based on the behavior distribution table Dst, calculate the discreteness change of the multi-agent behavior distribution to form a discreteness sequence Div, and extract the candidate time positions where behavioral consistency changes abruptly in the discreteness sequence Div to form a consistency breakdown candidate point Cbr. S4. Based on the discrete sequence Div, construct the consistency breakdown criterion Brk, determine whether the consistency breakdown candidate point Cbr satisfies the behavioral consistency breakdown condition, and generate the failure evaluation result Evl to represent the controllable state of the planning scheme. S5. Based on the failure assessment result Evl, perform simulation optimization on the urban planning scheme and output the planning optimization output Opt to represent the optimization result.

2. The urban planning optimization method based on multi-agent simulation according to claim 1, characterized in that: S1 includes S11; S11. In an urban simulation environment, the behavior of various intelligent agents participating in the simulation is monitored, and the original behavior records generated by the multiple intelligent agents during the simulation operation are collected. The original behavior record includes information on the agent's spatial location change, the time of the behavior, the duration of the behavior, and the behavior type identifier. The collected raw behavior records are standardized according to a unified time reference and spatial coordinate system, and merged according to the agent identifier to form a behavior record set Rec representing the original behavior state of multiple agents.

3. The urban planning optimization method based on multi-agent simulation according to claim 2, characterized in that: S1 further includes S12; S12. Based on the behavior record set Rec, according to the preset time continuity rules and spatial coherence rules, the continuous behavior records of the same agent are associated and reconstructed to generate a behavior trajectory that reflects the continuous behavior evolution process of the agent within the simulation cycle. The reconstructed behavioral trajectories of each agent are organized and summarized according to time windows and spatial units to form a set of behavioral trajectories Beh, which represents the behavioral response characteristics of multiple agents under the planning scheme.

4. The urban planning optimization method based on multi-agent simulation according to claim 3, characterized in that: S2 includes S21; S21. Based on the set of behavioral trajectories Beh, the behavioral trajectories of multiple agents are grouped and statistically analyzed according to a preset statistical time window and spatial analysis unit. The behavioral results of multiple agents within the same time window and the same spatial analysis unit are aggregated to form a distribution description for representing the structure of group behavior. The behavior distribution table Dst includes the following distribution contents: the distribution structure of multi-agent behavior trajectories in the activity duration dimension, the distribution structure of multi-agent behavior trajectories in the path selection feature dimension, and the distribution structure of multi-agent behavior trajectories in the activity chain integrity dimension.

5. The urban planning optimization method based on multi-agent simulation according to claim 4, characterized in that: S3 includes S31; S31. Based on the behavior distribution table Dst, and according to the time sequence of the statistical time windows, calculate the dispersion of the following distribution contents for the multi-agent behavior distribution structure within each statistical time window: The distribution structure of multi-agent behavioral trajectories along the activity duration dimension; The distribution structure of multi-agent behavioral trajectories along the path selection feature dimension; The distribution structure of multi-agent behavioral trajectories along the activity chain integrity dimension; The discreteness calculation for each type of distribution structure is performed using a unified distribution consistency evaluation method. The distribution consistency evaluation method includes: within the same statistical time window, based on the overall dispersion of the behavioral values ​​of each agent in the corresponding distribution structure, evaluating the degree of deviation of the distribution structure from the centralized state, and obtaining the single discreteness result corresponding to the distribution structure. Subsequently, the dispersion results corresponding to the three types of distribution structures within the same statistical time window are synthesized to obtain a comprehensive dispersion result that represents the overall degree of differentiation of multi-agent behavior within the statistical time window. The comprehensive dispersion results corresponding to each statistical time window are arranged in chronological order to form a dispersion sequence Div that describes the evolution of the consistency of multi-agent behavior over time.

6. The urban planning optimization method based on multi-agent simulation according to claim 5, characterized in that: S3 further includes S32; S32. Based on the discreteness sequence Div, the discreteness results corresponding to adjacent statistical time windows are compared one by one according to the time order of the statistical time windows to obtain the continuous change relationship of discreteness with time. In the discrete sequence Div, when the discrete result corresponding to a certain statistical time window suddenly increases compared to the discrete result corresponding to the previous statistical time window, and the sudden increase continues to exist in at least one subsequent adjacent statistical time window, the statistical time window is marked as a candidate time position where the behavioral consistency changes abruptly. Meanwhile, in the discrete sequence Div, when the discrete result shows a unidirectional expansion trend in multiple consecutive statistical time windows, that is, when the discrete result corresponding to the subsequent statistical time window is not lower than the discrete result corresponding to the previous statistical time window, the starting statistical time window of the continuous expansion trend is marked as the candidate time position where the behavioral consistency changes abruptly. The candidate time locations will be aggregated to form a consistency breakdown candidate point Cbr, which represents the potential risk of consistency breakdown in the behavior of multiple agents. The sudden increase state refers to an abnormal jump in the discrete result corresponding to a certain statistical time window, which exceeds a preset proportion range relative to the normal change range of the discrete sequence Div in the stable phase. The unidirectional expansion trend refers to the evolutionary state in which the dispersion result shows a continuous expansion without any pullback within multiple consecutive statistical time windows.

7. The urban planning optimization method based on multi-agent simulation according to claim 6, characterized in that: S4 includes S41; S41. Based on the discrete sequence Div, select a continuous statistical time window located in the stage where the behavioral consistency is in a stable state as the benchmark analysis segment. Within the benchmark analysis section, the level of change in the statistical dispersion results is used to determine the dispersion reference range corresponding to the stable phase. Based on the aforementioned dispersion reference range, a consistency breakdown criterion Brk is constructed to determine whether behavioral consistency has substantially broken down. The consistency breakdown criterion Brk is used to determine that when the discrete sequence Div continuously deviates from the discrete reference range and remains in an expanding state at the time position corresponding to the consistency breakdown candidate point Cbr, the behavioral consistency is determined to enter a breakdown state.

8. The urban planning optimization method based on multi-agent simulation according to claim 7, characterized in that: S4 also includes S42; S42. For the consistency breakdown candidate point Cbr, call the consistency breakdown criterion Brk to determine whether the corresponding statistical time window meets the behavioral consistency breakdown condition. When the consistency breakdown criterion Brk is met, the corresponding planning scheme is determined to enter an uncontrollable state within the statistical time window and is marked as a failure state in the failure evaluation result Evl; when the consistency breakdown criterion Brk is not met, the corresponding planning scheme is determined to remain in a controllable state within the statistical time window and is marked as a controllable state in the failure evaluation result Evl.

9. The urban planning optimization method based on multi-agent simulation according to claim 8, characterized in that: S5 includes S51; S51. Based on the failure assessment result Evl, analyze the operation status of the urban planning scheme during the simulation process, and identify the time position that leads the planning scheme into an uncontrollable state and the corresponding behavioral consistency collapse characteristics. For the planning schemes marked as uncontrollable in the failure assessment result Evl, perform structural adjustments at the simulation level, and re-execute steps S1 to S4 to obtain the adjusted failure assessment result Evl. The failure assessment result Evl corresponding to the adjusted planning scheme is compared with the failure assessment result Evl of the original planning scheme. The planning scheme in which the behavioral consistency collapse is eliminated is selected as the optimization result, and the planning optimization output Opt is output to represent the optimization effect of the planning scheme. The planning optimization output Opt includes the controllable state determination result of the planning scheme, the change of the behavior consistency collapse point, and the corresponding discrete sequence Div.

10. A city planning optimization system based on multi-agent simulation, applied to the city planning optimization method based on multi-agent simulation as described in any one of claims 1 to 9, characterized in that: It includes a simulation data acquisition module, a distributed extraction module, a distributed discrete computing module, a consistency judgment module, and an optimization decision module; The simulation data acquisition module collects behavioral data of multiple agents in an urban simulation environment and constructs a set of behavioral trajectories Beh representing the behavioral responses of planning schemes. The distribution extraction module constructs a behavior distribution table Dst that reflects the behavior distribution structure of multiple agents based on the behavior trajectory set Beh; The distributed discrete computing module calculates the discreteness change of the multi-agent behavior distribution according to the behavior distribution table Dst, forming a discreteness sequence Div, and extracts the candidate time positions where behavioral consistency changes abruptly in the discreteness sequence Div, forming a consistency collapse candidate point Cbr. The consistency judgment module constructs a consistency breakdown criterion Brk based on the discrete sequence Div, determines whether the consistency breakdown candidate point Cbr satisfies the behavioral consistency breakdown condition, and generates a failure assessment result Evl to represent the controllable state of the planning scheme. The optimization decision module performs simulation optimization of the urban planning scheme based on the failure assessment result Evl, and outputs the planning optimization output Opt to represent the optimization result.

Images