Software planning aid decision-making method and system based on space-time big data

Abstract

The invention relates to the field of intelligent traffic, and discloses a space-time big data-based software planning auxiliary decision-making method and system, and the method comprises the following steps: fusing mobile phone signaling, a vehicle GPS track, weather and traffic event data, constructing a dynamically updated space-time knowledge graph, and defining a space-time association relationship among a demand point, a connection station and a traffic event node; establishing a government, enterprise and passenger three-party asymmetric Nash negotiation model based on the knowledge graph, dynamically adjusting bargaining capability parameters in combination with real-time traffic events, and solving a game equilibrium solution; constructing a mixed integer collaborative optimization model by taking the equilibrium solution as a constraint condition, and solving a connection station layout and departure frequency scheme by adopting a two-stage algorithm; dynamically correcting a target function weight parameter according to the passenger flow prediction error through a PID closed-loop feedback mechanism; and finally, outputting a visual decision scheme. According to the method, the real-time decision-making efficiency and the scheme robustness in a large-scale urban road network scene are improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a software planning auxiliary decision-making method and system based on spatiotemporal big data. Background Art

[0002] Urban transit connection planning, as a crucial component of public transportation system optimization, must balance multi-dimensional objectives, including government policy guidance, enterprise operational efficiency, and passenger travel experience. Traditional approaches often rely on static traffic survey data and empirical models for resource allocation, struggling to effectively integrate real-time dynamic data sources like mobile phone signaling and vehicle trajectories. This results in inaccurate demand forecasts and fragmented representation of spatiotemporal characteristics. Existing technologies, particularly for the collaborative optimization of multi-agent interests, often employ fixed-weight linear programming or single-agent-led decision-making models. These techniques fail to quantify the dynamic game-playing relationships among government, enterprises, and passengers, and can easily lead to imbalanced resource allocation.

[0003] Furthermore, conventional optimization models are often based on offline calculations and lack closed-loop response mechanisms to real-time passenger flow fluctuations and unexpected traffic incidents. This results in poorly robust planning schemes and makes them difficult to adapt to the dynamic changes of large-scale urban road networks. At the decision-making support level, existing visualization tools are mostly limited to two-dimensional display, lacking temporal and spatial correlation analysis and interpretable representation of multi-objective coordination states, hindering the efficient translation of complex planning results into actionable strategies.

[0004] The above-mentioned defects jointly restrict the scientificity and real-time nature of urban traffic connection planning, and urgently need to be broken through through systematic innovation of data fusion, dynamic modeling and intelligent decision-making technology. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a software planning auxiliary decision-making method and system based on spatiotemporal big data, which solves the technical problems in traditional transportation connection planning methods, such as insufficient fusion of multi-source spatiotemporal data, difficulty in quantitative modeling of dynamic game of interests among multiple subjects, poor adaptability to real-time environments, and weak interpretability of complex planning results.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a software planning decision-making assistance method based on spatiotemporal big data, comprising the following steps:

[0007] S1, integrating multi-source spatiotemporal data to construct a dynamically updated spatiotemporal knowledge graph, which contains the spatiotemporal relationships among demand points, connecting stations, and traffic event nodes;

[0008] S2, based on the dynamically updated spatiotemporal knowledge graph, establishing an asymmetric Nash negotiation model among government entities, business entities, and passenger entities, and solving the game equilibrium solution by dynamically adjusting bargaining power parameters, wherein the bargaining power parameters are adaptively updated according to real-time traffic event data;

[0009] S3, using the game equilibrium solution as a constraint to construct a mixed integer collaborative optimization model, and solving the docking station layout and departure frequency plan through a two-stage algorithm, wherein the two-stage algorithm includes a phased search strategy for screening the range of candidate solutions based on the game equilibrium solution;

[0010] S4, generating a target weight adjustment signal based on the real-time passenger flow prediction error, and dynamically correcting the objective function parameters in the mixed integer collaborative optimization model;

[0011] S5 outputs a visual decision-making plan including a heat map of the docking station layout, a dynamic curve of the departure frequency, and a comparison of the utility of multiple agents.

[0012] Preferably, in step S1, the step of fusing multi-source spatiotemporal data to construct a dynamically updated spatiotemporal knowledge graph includes:

[0013] S11, collects mobile phone signaling data, vehicle GPS trajectory, weather data and traffic event data;

[0014] S12, fill in missing spatiotemporal data by spatiotemporal kriging interpolation;

[0015] S13, define demand points, connecting stations, traffic event nodes and their spatiotemporal correlation edges, and build a dynamically updated knowledge graph, where the weight of the spatiotemporal correlation edge is calculated based on the time decay factor w. ij (t) = e -δΔt Update, w ij (t) is the edge weight, Δt is the data time difference, and δ is the decay coefficient obtained from historical data training.

[0016] Preferably, in step S2, the step of establishing an asymmetric Nash negotiation model among the government entity, the enterprise entity, and the passenger entity includes:

[0017] S21, define the utility function U of the government entity G =α1S cover -α2C cong , the utility function U of the enterprise entity E =β1R income -β2C oper , the utility function U of the passenger P =-γT total ;

[0018] Among them, S cover is the service coverage, C cong is the congestion index, R income is operating income, C oper is the operating cost, T total To complete a total travel time, UG ,U E ,U P are the utility functions of the government, enterprise and passenger entities respectively; α1 and α2 are the weight coefficients of the influence of the preset service coverage and congestion index on the government utility; β1 and β2 are the weight coefficients of the influence of the preset operating income and operating cost on the enterprise utility; γ is the weight coefficient of the influence of the preset total time to complete a trip on the passenger utility.

[0019] Preferably, in step S2, the step of solving the game equilibrium solution by dynamically adjusting the bargaining power parameters includes:

[0020] S22, based on real-time traffic event data, calculate the bargaining power parameters θ of the government, enterprises, and passengers respectively G (t),θ E (t),θ P (t), where:

[0021] Passenger bargaining power parameter θ P (t) is calculated as follows:

[0022]

[0023] Among them, N event (t) is the number of traffic events in the current period, is the average number of traffic incidents in the same period of history, η P is the passenger bargaining sensitivity coefficient, The base value of passengers’ bargaining power;

[0024] The government's bargaining power parameter θ G (t) is calculated as follows:

[0025]

[0026] in, Indicates the current connection coverage rate With target threshold The difference, η G is the government bargaining sensitivity coefficient, The base value of the government's bargaining power;

[0027] The bargaining power parameter θ of the enterprise E (t) is calculated as follows:

[0028]

[0029] Among them, C oper is the enterprise operating cost, R income is the enterprise operating income, η E is the enterprise bargaining sensitivity coefficient, The base value of the enterprise's bargaining power;

[0030] S23, the bottom-line utilities of the three parties, namely the government, enterprises and passengers, are Based on the utility function U of each party G ,U E ,U P And the bargaining power parameter θ calculated in real time G (t),θ E (t),θ P (t) Construct the following Nash negotiation objective function:

[0031]

[0032] in, They represent the utility gains of the government, enterprises, and passengers relative to their bottom-line utilities respectively;

[0033] S24, solving the problem through an iterative optimization algorithm. When the utility value change rate of the three parties during the iteration process is less than the preset convergence threshold, the algorithm is judged to have converged, and the solution obtained at this time is output as the game equilibrium solution.

[0034] Preferably, in step S3, the step of using the game equilibrium solution as a constraint condition to construct a mixed integer collaborative optimization model includes:

[0035] S31, defining decision variables of the mixed integer collaborative optimization model, wherein the decision variables include:

[0036] Binary decision variable z j , when z j =1 means to choose to build connecting station j, when z j =0 means not to choose to build connecting station j;

[0037] Positive integer decision variable f k , represents the departure frequency of planned bus route k;

[0038] Continuous decision variable x ij , represents the proportion of travel demand of demand point i allocated to connecting station j;

[0039] S32: Extract or determine the theoretical optimal allocation ratio of each demand point i to each potential connecting station j from the obtained game equilibrium solution, denoted as

[0040] S33, constructing the objective function of the mixed integer collaborative optimization model to minimize the comprehensive cost, the comprehensive cost is the weighted total passenger travel time, the total operating cost of the enterprise, and the actual allocation ratio and the theoretical optimal allocation ratio The deviation between them is composed of the expression of the objective function:

[0041]

[0042] Among them, T total Represented by the decision variable z j ,f k ,x ij The total travel time of all passengers in the system is determined by C oper Represented by the decision variable z j ,f k ,x ij The total operating cost of the enterprise determined by ω1 and ω2 are the weight parameters of the total travel time of the passengers and the total operating cost of the enterprise respectively; x is the actual allocation ratio for the penalty ij Deviation from the theoretical optimal distribution ratio Deviation penalty coefficient; represents the actual allocation ratio x from all demand points i to all connecting stations j ij Compared with the theoretical optimal allocation ratio The total absolute deviation between

[0043] S34, setting constraints for the mixed integer collaborative optimization model, wherein the constraints at least include:

[0044] Demand satisfaction constraint: The allocation ratio of each demand point i satisfies ∑ j x ij =1;

[0045] Connecting station capacity constraint: the total passenger flow of connecting station j does not exceed its capacity limit;

[0046] Line departure frequency constraint: The departure frequency meets the preset range;

[0047] Game equilibrium solution coordination constraint: the absolute value of the deviation between the demand allocation ratio and the equilibrium solution does not exceed the preset threshold.

[0048] Preferably, in step S3, the step of solving the connection station layout and departure frequency plan by a two-stage algorithm includes:

[0049] When screening candidate solutions in the first stage, based on the constraints of the game equilibrium solution coordination constraints, a set of candidate solutions that meet the conditions is generated through a greedy search algorithm;

[0050] In the second stage of exact solution, the top M solutions with the smallest objective function value are selected from the generated candidate solution set as the initial solution, and the decision variable z in each initial solution is j and f kThe mixed integer collaborative optimization model is iteratively optimized and solved by the branch and bound method until the final connection station layout and departure frequency plan is obtained, which can make the objective function reach the optimal state or meet the preset optimization termination conditions.

[0051] Preferably, in step S4, the step of generating a target weight adjustment signal based on the real-time passenger flow prediction error and dynamically correcting the objective function parameters in the mixed integer collaborative optimization model includes:

[0052] S41, real-time monitoring of the passenger flow prediction error e of each connecting station j in the current period t j (t), the passenger flow prediction error e j (t) is calculated by the following formula:

[0053]

[0054] in, is the predicted passenger flow of connecting station j in time period t, is the actual observed passenger flow of connecting station j in time period t;

[0055] S42, based on the real-time passenger flow prediction error e of all J connecting stations j (t), calculate the average passenger flow prediction error in time period t The calculation formula is:

[0056]

[0057] Where J is the total number of connection sites in the system;

[0058] S43, using a PID controller, based on the calculated average passenger flow prediction error The passenger total travel time weight parameter ω1 and the enterprise operating cost weight parameter ω2 used in the objective function of the mixed integer collaborative optimization model are dynamically calculated and updated in the following manner:

[0059] The updated value ω1(t) of the passenger total travel time weight parameter in time period t is calculated by the following formula:

[0060]

[0061] The updated value of the enterprise operating cost weight parameter in time period t, ω2(t), is calculated by the following formula:

[0062]

[0063] in, is the preset base value of the passenger's total travel time weight parameter; K p ,K i ,Kd are the proportional, integral, and differential coefficients of the PID controller; is the cumulative integral value of the average passenger flow prediction error from the preset integration start time t0 to the current time period t, where e(τ) represents the average passenger flow prediction error at time τ; is the rate of change of the average passenger flow prediction error in the current period t; is the preset base value of the enterprise operating cost weight parameter; γ1 is the preset weight coupling coefficient;

[0064] S44, the updated passenger total travel time weight parameter ω1(t) and the enterprise operating cost weight parameter ω2(t) are fed back in real time to the objective function of the mixed integer collaborative optimization model to replace the original corresponding weight parameters in the subsequent solution or iterative adjustment process of the model.

[0065] Preferably, the method further comprises the following steps:

[0066] Provide a manual interface to adjust the base value parameters of the bargaining power of governments, businesses, and passengers.

[0067] The present invention also provides a software planning decision-making support system based on spatiotemporal big data, comprising:

[0068] Data fusion module, used to fuse multi-source spatiotemporal data to build a dynamically updated spatiotemporal knowledge graph;

[0069] A game modeling module is used to establish an asymmetric Nash negotiation model among government entities, business entities, and passenger entities based on the dynamically updated spatiotemporal knowledge graph, and to solve the game equilibrium solution by dynamically adjusting bargaining power parameters;

[0070] An optimization solution module uses the game equilibrium solution as a constraint to construct a mixed integer collaborative optimization model and solve the connection station layout and departure frequency plan through a two-stage algorithm;

[0071] a feedback control module, generating a target weight adjustment signal based on the real-time passenger flow prediction error, and dynamically correcting the objective function parameters in the mixed integer collaborative optimization model;

[0072] The decision output module outputs a visual decision plan including a heat map of the docking station layout, a dynamic curve of the departure frequency, and a comparison of the utility of multiple agents, and supports manual intervention.

[0073] Preferably, the optimization solution module includes:

[0074] A candidate solution screening unit, used to screen the range of feasible solutions based on the game equilibrium solution constraints;

[0075] A branch and bound solving unit is used to accurately solve integer variables within a range of candidate solutions.

[0076] The present invention provides a software planning decision-making support method and system based on spatiotemporal big data. It has the following beneficial effects:

[0077] 1. This invention builds a dynamically updated spatiotemporal knowledge graph by fusing multi-source spatiotemporal data, integrating multi-dimensional data such as mobile phone signaling, vehicle GPS, meteorological and traffic events, using spatiotemporal kriging interpolation to fill in missing data, and defining dynamically decaying spatiotemporal correlation edges. This solves the problems of data dispersion and poor timeliness in traditional planning, and significantly improves the accuracy of traffic demand forecasting and resource matching.

[0078] 2. This invention establishes an asymmetric Nash negotiation model among the government, enterprises, and passengers, dynamically adjusts bargaining power parameters based on real-time traffic event data, and transforms the multi-agent interest game into an exponentially weighted product optimization problem. This achieves a balanced decision-making process among policy goals, enterprise benefits, and passenger experience, and avoids the planning imbalance caused by a single dominant entity in traditional methods.

[0079] 3. The present invention constructs a mixed integer collaborative optimization model and designs a two-stage solution algorithm. In the first stage, the range of candidate solutions is screened based on the game equilibrium solution. In the second stage, the branch and bound method is used to refine the search. While ensuring the quality of the solution, the computational complexity is reduced, effectively solving the problem of low efficiency of integer programming solution in large-scale road network scenarios.

[0080] 4. The present invention dynamically adjusts the objective function weight through a PID closed-loop feedback mechanism based on real-time passenger flow prediction errors, uses proportional, integral, and differential terms to respond to current errors, cumulative errors, and future fluctuations, respectively, and combines reinforcement learning to optimize PID coefficients, thereby enhancing the model's adaptability to sudden changes in passenger flow and avoiding planning rigidity caused by static weights.

[0081] 5. The present invention outputs multi-dimensional visual decision-making solutions, displays the heat map of docking stations, the dynamic curve of departure frequency, and the multi-agent utility comparison panel in a spatiotemporal linkage, and integrates a manual intervention interface and data drilling function to make the results of complex mathematical models interactive and interpretable, providing decision makers with an integrated support tool from macro-strategy to micro-adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 Schematic diagram of the method flow of the present invention;

[0083] Figure 2 Schematic diagram of the system structure of the present invention.

[0084] Among them, 10, data fusion module; 20, game modeling module; 30, optimization solution module; 40, feedback control module; 50, decision output module. DETAILED DESCRIPTION

[0085] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0086] The present invention provides a software planning auxiliary decision-making method based on spatiotemporal big data. The method constructs a knowledge graph by integrating multi-source dynamic data, establishes a multi-agent game model and a collaborative optimization framework, and realizes intelligent decision-making on the layout of docking stations and departure frequency.

[0087] Please see the attached Figure 1 The software planning decision-making assistance method based on spatiotemporal big data may include the following steps:

[0088] S1. Integrate multi-source spatiotemporal data to build a dynamically updated spatiotemporal knowledge graph;

[0089] S2. Based on the spatiotemporal knowledge graph, an asymmetric Nash negotiation model is established among the three parties: government, enterprises, and passengers.

[0090] S3. Using the game equilibrium solution as a constraint, a mixed integer collaborative optimization model is constructed to solve the connection station layout and departure frequency plan through a two-stage algorithm.

[0091] S4. Generate a target weight adjustment signal based on the real-time passenger flow prediction error and dynamically modify the objective function parameters in the mixed integer collaborative optimization model;

[0092] S5. Output visual decision plan.

[0093] The following is a detailed description of each step in the method of the present invention, which comprehensively explains the specific implementation principles, technical details and processes of each step.

[0094] In step S1, multi-source spatiotemporal data are integrated to build a dynamically updated spatiotemporal knowledge graph. By integrating heterogeneous data and using spatiotemporal modeling technology, high-precision spatiotemporal feature expression required for transportation connection planning is achieved. Specifically, the construction process of the spatiotemporal knowledge graph includes:

[0095] First, we collect multi-source heterogeneous spatiotemporal data as the foundational data source for knowledge graph construction. This multi-source spatiotemporal data includes mobile phone signaling data, vehicle GPS trajectory data, meteorological data, and traffic event data. Mobile phone signaling data is used to extract user travel trajectories and demand distribution, vehicle GPS trajectory data is used to analyze road traffic conditions and vehicle operation patterns, meteorological data is used to assess the impact of weather factors on traffic demand, and traffic event data includes the real-time location and impact range of traffic accidents and road construction.

[0096] Furthermore, the collected raw data is subjected to spatiotemporal alignment and missing value filling. Due to the lack of spatiotemporal data caused by signal loss or equipment failure during the data collection process, the spatiotemporal characteristics of the missing areas are estimated using the spatiotemporal kriging interpolation method. Specifically, the spatiotemporal kriging interpolation method generates continuous spatiotemporal field data by establishing a spatiotemporal covariance function and calculating the weighted relationship between the spatial distance and time interval between the spatiotemporal point to be filled and the adjacent known data points. Its interpolation formula is expressed as:

[0097]

[0098] Among them, z(s,t) represents the space-time coordinate (s i ,t i ) interpolation result, λ i is the weight coefficient, which is calculated by combining the spatiotemporal covariance function with the spatial distance attenuation factor δ. The attenuation factor is determined through historical data training and is used to quantify the attenuation rate of spatiotemporal correlation.

[0099] On this basis, knowledge graph nodes and their relationships are defined. Nodes include three types of entities: demand points, connecting stations, and traffic event nodes. Demand points are generated using a city grid system, with each grid center serving as a demand point. Node attributes include historical passenger flow, walking distance, and real-time demand forecasts. Connecting stations are candidate bus or shared transportation connection locations. Node attributes include station capacity, construction cost, and service coverage radius. Traffic event nodes record event type, occurrence time, and geofence range.

[0100] Furthermore, spatiotemporal edges are constructed between nodes to quantify the dynamic interaction between entities. These edges include service-reachable edges between demand points and connecting stations, and propagation-impact edges between traffic event nodes and affected areas. The weight of the service-reachable edge is dynamically updated based on the time decay factor, calculated as follows:

[0101] w ij (t) = e -δ·Δt

[0102] Among them, w ij(t) is the edge weight, Δt represents the data time difference, that is, the difference between the current time and the data collection time, and δ is the decay coefficient obtained through historical data training, which is used to characterize the decay rate of spatiotemporal correlation. The decay coefficient is set differently based on the spatiotemporal correlation pattern of different edge types. For example, the decay coefficient of traffic event-affected edges is larger than that of regular service-reachable edges to reflect the short-term impact of emergencies.

[0103] The dynamic update mechanism of the spatiotemporal knowledge graph includes two modes: periodic updates and event-triggered updates. Preferably, periodic updates are performed at fixed intervals (e.g., 15 minutes), globally refreshing node attributes and edge weights. Event-triggered updates, upon detecting a new traffic incident or sudden change in passenger flow, locally reconstruct the node and edge relationships in the affected area. This update mechanism ensures that the knowledge graph can reflect the dynamic changes of the transportation system in real time.

[0104] Finally, the spatiotemporal knowledge graph is stored using a combination of a graph database and a spatiotemporal index. Preferably, a Neo4j graph database is used to store the topology of node and edge relationships, while an R-tree index is used to quickly retrieve spatiotemporal attributes to support efficient data access for subsequent game models and optimization algorithms.

[0105] In step S2, an asymmetric Nash negotiation model among the government, enterprises, and passengers is established and the game equilibrium solution is solved. Through the adaptive adjustment mechanism of dynamic bargaining power parameters, the multi-agent interest game process is quantified into an iteratively optimized mathematical model. Specifically, the model construction and solution process includes:

[0106] First, based on the node attributes and associations extracted from the spatiotemporal knowledge graph, the utility functions of the three parties, namely the government, enterprises, and passengers, are defined. Among them, the utility function of the government entity takes public transportation service coverage and traffic congestion relief as core indicators, and is expressed as follows:

[0107] U G =α1S cover -α2C cong

[0108] Where U G is the utility function of the government entity, S cover represents the connection service coverage rate, which is calculated by the proportion of demand points covered by the connection stations; C cong is the traffic congestion index, which quantifies the degree of deviation between the average speed of the road section in the vehicle GPS trajectory data and the free flow speed; α1 and α2 are the weight coefficients of the impact of the preset service coverage and congestion index on government utility, respectively.

[0109] The utility function of the enterprise entity focuses on operating income and cost control, and its expression is defined as:

[0110] U E =β1R income -β2C oper

[0111] Among them, U E is the utility function of the enterprise entity, R income represents operating revenue, which is calculated by the proportion of passenger fares and demand distribution; C oper is the enterprise operating cost, including vehicle energy consumption, labor cost and equipment maintenance cost; β1 and β2 are the weight coefficients of the impact of the preset operating income and operating cost on the enterprise utility.

[0112] The utility function of the passenger subject is centered on minimizing travel time, and its expression is:

[0113] U P =-γT total

[0114] Among them, U P is the utility function of the passenger, T total is the total travel time of the passenger, including the time of walking from the demand point to the connecting station, waiting time and boarding time, which is calculated through the walking time, real-time departure frequency and line mileage data in the spatiotemporal knowledge graph; γ is the preset weight coefficient of the impact of the total travel time on the passenger utility.

[0115] Furthermore, dynamic bargaining power parameters are introduced to reflect the real-time influence changes of multiple subjects during the game process.

[0116] The bargaining power parameter is dynamically adjusted based on the real-time data of the external traffic environment and internal operation status. Specifically, the passenger bargaining power parameter θ P The calculation formula for (t) is:

[0117]

[0118] Where N event (t) is the number of traffic events in the current period, which is obtained by real-time parsing of traffic event nodes in the spatiotemporal knowledge graph; is the average number of traffic incidents in the same period of history; η P is the sensitivity coefficient, which is used to regulate the extent to which traffic events enhance the bargaining power of passengers. The base value of the bargaining power of passengers is set by the government based on the principle of equalization of public services, reflecting the level of protection of the basic rights and interests of the passenger group. For example, in cities where public transportation is more public welfare, Set to above 0.4. When the number of sudden traffic incidents increases, passengers’ demands for travel efficiency increase, resulting in θ P (t) Promotion.

[0119] Government bargaining power parameter θ G The calculation of (t) is related to the degree of achievement of policy goals, and its expression is:

[0120]

[0121] in, Indicates the difference between the current connection coverage rate and the preset target threshold; η G is the government-side sensitivity coefficient, which is used to amplify or reduce the impact of coverage deviation on bargaining power. The government's bargaining power base value is determined by the priority of public service goals set in the government's initial policy documents. For example, if the policy emphasizes coverage first, then Set it to a higher value (such as 0.5), otherwise it will be reduced. The specific value is calibrated by historical policy effect evaluation data. When the actual coverage rate is lower than the target threshold, the government will increase the θ G (t) Increase its decision weight in the game, forcing the model to prioritize meeting coverage requirements.

[0122] Enterprise bargaining power parameter θ E The calculation of (t) is based on the operating cost profit margin, which is expressed as:

[0123]

[0124] Where C oper With R income They are obtained from the enterprise utility function, reflecting the proportion of real-time operating costs to revenue; η E is the enterprise side sensitivity coefficient. = is the base value of the enterprise's bargaining power, which is set based on the cost-benefit ratio in the enterprise's initial operating strategy. If the enterprise tends to adopt a high-cost and high-quality operating model, then When the cost-profit ratio decreases, the enterprise increases θ E (t) Strengthen its bargaining position in the game and encourage the model to reduce the intensity of punishment for its cost constraints.

[0125] After completing the definition of the three-party utility function and dynamic bargaining parameters, an asymmetric Nash negotiation objective function is constructed to solve the game equilibrium solution.

[0126] The bottom-line utilities of the three parties, namely government, enterprises and passengers, are Based on the utility function U of each party G ,U E ,U P And the bargaining power parameter θ calculated in real time G (t),θ E (t),θP (t) Construct the following Nash negotiation objective function:

[0127]

[0128] in, The objective function transforms the multi-agent game process into an exponentially weighted product using dynamic bargaining parameters, giving agents with stronger bargaining power a higher weight in the solution space.

[0129] The solution process for the game equilibrium solution uses a sequential quadratic programming algorithm for iterative optimization. Preferably, the algorithm is initialized with the historical optimal solution as the initial guess value. In each iteration, the bargaining power parameters are updated and the objective function gradient is recalculated until the rate of change of the three-party utility is less than the preset threshold. The rate of change is calculated as follows:

[0130]

[0131] Among them, U (k) It represents the tripartite utility value of the kth iteration, and ∈ is the preset convergence threshold. When the convergence condition is met, the game equilibrium solution is output. as constraints for the subsequent mixed integer optimization model.

[0132] Through the dynamic bargaining parameter adjustment mechanism and iterative optimization algorithm, the model can respond to changes in the status of the transportation system in real time, balance the conflicts of interest among multiple entities, and provide a game equilibrium solution for connection planning that takes into account policy goals, corporate benefits and passenger experience.

[0133] In step S3, the game equilibrium solution is used as a constraint to construct a mixed integer collaborative optimization model and solve the docking station layout and departure frequency plan through a two-stage algorithm. By integrating game theory and mathematical programming methods, multi-objective collaborative decision-making is achieved. Specifically, the model construction and solution process includes:

[0134] First, define the decision variables of the mixed integer collaborative optimization model. The variables include binary decision variables, positive integer decision variables, and continuous decision variables:

[0135] Binary decision variable z j , when z j =1 means to choose to build connecting station j, when z j =0 means not to choose to build connecting station j;

[0136] Positive integer decision variable f k , represents the departure frequency of the planned bus route k, and its value range is preset according to the operational feasibility, such as the minimum departure interval and the maximum number of departures.

[0137] Continuous decision variable x ij , which represents the proportion of travel demand of demand point i allocated to connecting station j. The integrity of demand allocation is ensured by normalization.

[0138] On this basis, a multi-objective optimization objective function is constructed to minimize the comprehensive cost, which is composed of the weighted total passenger travel time, the total operating cost of the enterprise, and the actual allocation ratio and the theoretical optimal allocation ratio. The objective function consists of three parts: the total travel time of passengers, the operating cost of enterprises, and the penalty term for deviation of the game equilibrium solution. Its expression is:

[0139]

[0140] Where, T total Represented by the decision variable z j ,f k ,x ij The total travel time of all passengers in the system is determined by C oper Represented by the decision variable z j ,f k ,x ij The total operating cost of the enterprise determined by ω1 and ω2 are the weight parameters of the total travel time of the passengers and the total operating cost of the enterprise respectively; λ is the preset actual allocation ratio x for the penalty ij Deviation from the theoretical optimal distribution ratio Deviation penalty coefficient; represents the actual allocation ratio x from all demand points i to all connecting stations j ij Compared with the theoretical optimal allocation ratio The total absolute deviation between

[0141] Among them, T total The calculation formula is:

[0142]

[0143] Where, is the walking time from demand point i to site j, calculated using the road network data in the spatiotemporal knowledge graph; Indicates the frequency of departure based on f k Average waiting time for buses; is the travel time of route k, which is determined by the route mileage d k With the average vehicle speed v k calculate.

[0144] Enterprise operating costs C oper The calculation formula is:

[0145]

[0146] in, is the energy consumption cost per trip on route k, which is related to the vehicle type and route slope; and They are labor cost and maintenance cost, which are preset based on the enterprise operation data.

[0147] Furthermore, the deviation penalty term of the game equilibrium solution Transform the solution into a solvable form through linearization. Preferably, introduce auxiliary variables and Represent positive and negative deviations respectively, satisfying:

[0148]

[0149] And rewrite the penalty term as Ensure the linearity of the objective function.

[0150] The constraints of the above model include the following four categories:

[0151] Requirements satisfy constraints:

[0152]

[0153] Ensure that passengers at each demand point i are fully assigned to connecting stations.

[0154] Connection site capacity constraints:

[0155]

[0156] where Q j is the maximum service capacity of station j, to prevent operational risks caused by passenger overload.

[0157] Line departure frequency constraints:

[0158]

[0159] The feasible range is preset based on vehicle dispatching capabilities and road resource constraints.

[0160] Game equilibrium solution coordination constraints:

[0161]

[0162] Where ∈ is the preset deviation threshold, which ensures the theoretical consistency of the optimization result and the game equilibrium solution.

[0163] After the model is built, a two-stage algorithm is used to solve the problem. The first stage of the two-stage algorithm uses greedy search to quickly generate a set of candidate solutions. Specifically, it includes the following steps:

[0164] Candidate solution generation: game equilibrium solution Centered, within the deviation constraint Random perturbations within the range generate the initial solution set;

[0165] Screening strategy: Calculate the objective function value of each solution and retain the top N solutions with the smallest objective function value as the candidate solution set Preferably, N is dynamically adjusted according to computing resources to balance efficiency and solution quality.

[0166] The second stage uses the branch-and-bound method to conduct a refined search for candidate solutions. The specific process includes:

[0167] Branch operation: for the integer variable z in the candidate solution j With f k , give priority to the departure frequency f k Perform bifurcation, generate child nodes and calculate the optimal lower bound LB of the relaxed model;

[0168] Pruning rule: If the lower bound LB of a node exceeds the upper bound UB of the current global optimal solution, the branch is pruned to reduce the search space;

[0169] Termination condition: When all branches are processed or the preset maximum number of iterations is reached, the optimal docking station layout plan is obtained to achieve the optimal objective function or meet the preset optimization termination conditions. and departure frequency plan

[0170] This two-stage algorithm significantly reduces computational complexity while maintaining solution quality, making it particularly suitable for real-time decision-making in large-scale urban road network scenarios. The generation and selection of candidate solution sets effectively narrows the search space, while the branch-and-bound method ensures the global optimality of the final solution through a systematic branching and pruning strategy.

[0171] Finally, the optimization results are matched and verified against real-time data in the spatiotemporal knowledge graph. Preferentially, solutions that exceed capacity constraints or the feasible range of departure frequencies are retroactively corrected to ensure the physical feasibility of the output plan. The mixed-integer collaborative optimization model and its solution algorithm form a closed-loop technology from theoretical equilibrium solutions to actual operational plans, providing highly robust decision support for multi-objective shuttle planning.

[0172] In step S4, a target weight adjustment signal is generated based on the real-time passenger flow prediction error and the objective function parameters in the mixed integer collaborative optimization model are dynamically modified. The adaptive optimization of the model parameters is achieved through a closed-loop feedback control mechanism. Specifically, the dynamic weight adjustment process includes:

[0173] First, the passenger flow prediction error e of each connecting station j in the current period t is monitored in real time.j (t), passenger flow prediction error e j (t) is calculated by the following formula:

[0174]

[0175] in, is the passenger flow of station j in time period t obtained by predicting the historical passenger flow data, weather factors and traffic events in the spatiotemporal knowledge graph; The actual passenger flow is calculated in real time using mobile phone signaling data and vehicle GPS trajectories. The prediction model is generated using a time series analysis algorithm (such as ARIMA or LSTM), and the prediction results are dynamically adjusted based on the spatiotemporal relationships in the knowledge graph.

[0176] Furthermore, the global average passenger flow prediction error is calculated to reflect the overall prediction deviation level. j (t), calculate the average passenger flow prediction error in time period t The calculation formula is:

[0177]

[0178] Where J is the total number of connecting stations currently in operation. As the input signal of the PID controller, it is used to drive the dynamic adjustment of the objective function weight parameters.

[0179] On this basis, the average passenger flow prediction error calculated Dynamically calculate and update the passenger total travel time weight parameter ω1 and the enterprise operating cost weight parameter ω2 used in the objective function of the mixed integer collaborative optimization model. The calculation formula is:

[0180]

[0181] Where, is the preset base value of the passenger's total travel time weight parameter; K p ,K i ,K d are the proportional, integral, and differential coefficients of the PID controller; is the cumulative integral value of the average passenger flow prediction error from the preset integration start time t0 to the current time period t, where e(τ) represents the average passenger flow prediction error at time τ; is the rate of change of the average passenger flow prediction error in the current period t; is the preset base value of the enterprise operating cost weight parameter; γ1 is the preset weight coupling coefficient.

[0182] The design logic of the PID controller includes the following core elements:

[0183] Proportional term: quickly responds to real-time prediction errors and adjusts weights directly based on the current error magnitude;

[0184] Integral term: eliminates historical cumulative errors and solves the weight shift problem caused by long-term forecast deviations;

[0185] Differential term: suppress future error fluctuations and adjust weight parameters in advance by predicting the error change rate.

[0186] After the weight parameters are updated, the adjusted ω1(t) and ω2(t) are fed back to the objective function of the mixed integer collaborative optimization model in real time. Preferably, the feedback mechanism triggers the model to be re-solved at a fixed time interval (e.g., 30 minutes), or is triggered dynamically according to the error threshold condition. When the average error is detected, When the preset threshold is exceeded, the weight parameters are immediately updated and the optimization algorithm is rerun to ensure the model's ability to respond quickly to sudden passenger flow fluctuations.

[0187] The dynamic weight adjustment process and the optimization model solution process form a closed-loop control loop. Specifically, the connection station layout and departure frequency scheme obtained by the model solution will affect the actual passenger flow distribution, thereby changing the prediction error e in the subsequent period. j (t), and the error data drives the update of the weight parameters again, forming a closed-loop feedback loop of "prediction-error-adjustment-optimization." Through this closed-loop mechanism, the model can adapt to changes in the external environment and continuously approach the dynamic equilibrium state of multi-objective optimization.

[0188] Finally, the weight parameter adjustment process supports manual intervention interface. Preferably, the decision maker can manually modify the PID coefficient K p ,K i ,K d Or the weight coupling coefficient γ to adapt to the operational strategy requirements in special scenarios (such as large-scale events and extreme weather). The human intervention data will be used as prior knowledge for the reinforcement learning algorithm to optimize the subsequent PID parameter self-learning process.

[0189] Through the above technical solution, step S4 realizes the dynamic optimization and closed-loop control of the objective function parameters, so that the mixed integer collaborative optimization model can effectively deal with the uncertainty of passenger flow prediction and improve the robustness and real-time performance of the connection planning scheme.

[0190] In step S5, a visualization decision plan is output, including a heat map of the docking station layout, a dynamic curve of the departure frequency, and a comparison of the utility of multiple agents. Through multi-dimensional data fusion and interactive visualization technology, the complex planning results are transformed into actionable decision support information. Specifically, the generation and presentation process of the visualization plan includes:

[0191] First, a heat map of the connection station layout is generated to visually show the matching relationship between demand distribution and connection service. The construction of the heat map is based on the station selection variable z output by the optimization model. j and the demand allocation variable x ij , the specific method is:

[0192] Demand thermal layer: With the geographical location of demand point i as the center, according to the allocation ratio x ij The demand density distribution is generated by overlaying, and the color mapping adopts the HSV model, where the hue represents the demand intensity and the saturation reflects the service coverage gap;

[0193] Site coverage: The selected connection site j(z j =1) is superimposed on the heat map, with circular marks indicating the site location, radius and site capacity Q j Positive correlation, the color depth indicates the real-time passenger flow load rate

[0194] The dynamic update of the heat map is synchronized with the refresh cycle of the spatiotemporal knowledge graph. Preferably, when a traffic incident or a sudden change in passenger flow is detected, a local re-rendering of the heat map is triggered to highlight the supply and demand imbalance in the affected area.

[0195] Furthermore, a dynamic curve of train departure frequency is drawn to show the time sequence change of the line operation plan. The horizontal axis of the curve is the time axis (24-hour cycle), and the vertical axis is the train departure frequency f k , generated by the following data fusion method:

[0196] Historical baseline: The historical average of departure frequencies during the same period extracted from the spatiotemporal knowledge graph, represented by a dotted line;

[0197] Real-time optimization line: based on the optimization results of the current period It is represented by a solid line and a rolling forecast is made for future periods to generate trend bands with confidence intervals;

[0198] Abnormal marking point: When the adjustment range of the departure frequency of a certain line exceeds the preset threshold (for example, ±30%), a warning mark is added at the corresponding time point, and the cause of the constraint triggering in the optimization model (such as capacity exceeding the limit or the equilibrium solution deviation is too large) is displayed.

[0199] In the construction of the multi-agent utility comparison panel, a mixed chart format is used to simultaneously display the utility changes and bargaining power parameter dynamics of the three parties: government, enterprises, and passengers. Specifically, it includes:

[0200] Utility value bar chart: Use different colors to distinguish U G ,U E ,U PThe column height indicates the utility value of the current period, and the background color transparency is related to the convergence state of the game equilibrium solution (lower transparency indicates worse convergence);

[0201] Bargaining Power Trend Line: Plot θ on the secondary axis G (t),θ E (t),θ P The change curve of (t) has the same line type and color as the main body, reflecting the real-time impact of the external environment on the bargaining position of each party;

[0202] Interactive drilldown: Supports clicking on a specific time period to view detailed data, including constraint violation records, weight parameter adjustment history, and a list of associated traffic events.

[0203] The visual decision solution uses WebGL technology to achieve high-performance rendering and integrates a human intervention interface. Preferably, the interface provides the following interactive functions:

[0204] Heatmap parameter adjustment: allows decision makers to manually adjust the color mapping range, site marker size, and data filtering conditions to focus on key areas;

[0205] Manual correction of departure frequency: In the departure frequency dynamic curve interface, you can drag and drop the optimization result data points to force the departure frequency of a specific period and trigger the model to re-solve to evaluate the impact;

[0206] Utility weight reset: In the multi-agent utility panel, slider controls are provided to temporarily override the ω1(t) and ω2(t) parameters output by the PID controller to observe the utility changes under different weight combinations.

[0207] Finally, the visualization solution achieves multi-view linkage through a spatiotemporal indexing mechanism. When a user selects a connecting station or time interval, the associated thermal layer, departure frequency curve segment, and utility comparison data point are automatically highlighted. The underlying data details (such as station construction costs and passenger flow forecast model error distribution) are then displayed through a knowledge graph query interface. This linkage mechanism is implemented through event bus technology, ensuring consistent data updates and low-latency responses.

[0208] Through the above technical solution, step S5 converts the complex mathematical model output into a hierarchical, interactive visual expression, allowing decision makers to quickly identify weak links in connection planning, assess the balance of interests among multiple entities, and make dynamic strategy adjustments based on real-time data, thereby realizing data-driven refined traffic management.

[0209] In general, the present invention constructs a dynamically updated spatiotemporal knowledge graph by integrating multi-source spatiotemporal data, establishes an asymmetric Nash negotiation model among the government, enterprises, and passengers to solve the game equilibrium solution, and uses it as a constraint condition to construct a mixed integer collaborative optimization model. The two-stage algorithm is combined to solve the connection station layout and departure frequency plan, and the objective function weight parameters are dynamically adjusted through the PID closed-loop feedback mechanism based on the real-time passenger flow prediction error. Finally, a visual decision-making plan that integrates heat maps, dynamic curves, and multi-agent utility comparison is output, realizing data-driven multi-objective collaborative optimization and dynamic adaptive decision-making of connection planning.

[0210] The software planning decision support system based on spatiotemporal big data described below and the software planning decision support method based on spatiotemporal big data described above can refer to each other.

[0211] Please see the attached Figure 2 The present invention also provides a software planning decision-making support system based on spatiotemporal big data, including:

[0212] The data fusion module 10 is used to fuse multi-source spatiotemporal data to build a dynamically updated spatiotemporal knowledge graph;

[0213] A game modeling module 20 is configured to establish an asymmetric Nash negotiation model among government entities, business entities, and passenger entities based on the dynamically updated spatiotemporal knowledge graph, and to solve a game equilibrium solution by dynamically adjusting bargaining power parameters;

[0214] The optimization solution module 30 uses the game equilibrium solution as a constraint to construct a mixed integer collaborative optimization model and solve the connection station layout and departure frequency plan through a two-stage algorithm;

[0215] A feedback control module 40 generates a target weight adjustment signal based on the real-time passenger flow prediction error, and dynamically modifies the objective function parameters in the mixed integer collaborative optimization model;

[0216] The decision output module 50 outputs a visual decision plan including a heat map of the docking station layout, a dynamic curve of the departure frequency, and a comparison of the utility of multiple agents, and supports manual intervention.

[0217] The device of this embodiment can be used to execute the above method embodiment, and its principles and technical effects are similar, so they will not be repeated here.

[0218] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A software planning decision-making support method based on spatiotemporal big data, characterized by: The following steps are involved: S1, integrating multi-source spatiotemporal data to construct a dynamically updated spatiotemporal knowledge graph, which contains the spatiotemporal relationships among demand points, connecting stations, and traffic event nodes; S2, based on the dynamically updated spatiotemporal knowledge graph, establishing an asymmetric Nash negotiation model among government entities, business entities, and passenger entities, and solving the game equilibrium solution by dynamically adjusting bargaining power parameters, wherein the bargaining power parameters are adaptively updated according to real-time traffic event data; S3, using the game equilibrium solution as a constraint to construct a mixed integer collaborative optimization model, and solving the docking station layout and departure frequency plan through a two-stage algorithm, wherein the two-stage algorithm includes a phased search strategy for screening the range of candidate solutions based on the game equilibrium solution; S4, generating a target weight adjustment signal based on the real-time passenger flow prediction error, and dynamically correcting the objective function parameters in the mixed integer collaborative optimization model; S5 outputs a visual decision-making plan including a heat map of the docking station layout, a dynamic curve of the departure frequency, and a comparison of the utility of multiple agents.

2. The software planning decision-making support method based on spatiotemporal big data according to claim 1 is characterized in that: In step S1, the step of fusing multi-source spatiotemporal data to construct a dynamically updated spatiotemporal knowledge graph includes: S11, collects mobile phone signaling data, vehicle GPS trajectory, weather data and traffic event data; S12, fill in missing spatiotemporal data by spatiotemporal kriging interpolation; S13, define demand points, connecting stations, traffic event nodes and their spatiotemporal correlation edges, and build a dynamically updated knowledge graph, where the weight of the spatiotemporal correlation edge is calculated based on the time decay factor w. ij (t) = e -δΔt Update, w ij (t) is the edge weight, Δt is the data time difference, and δ is the decay coefficient obtained from historical data training.

3. The software planning decision-making support method based on spatiotemporal big data according to claim 1 is characterized in that: In step S2, the steps of establishing an asymmetric Nash negotiation model among the government, the enterprise, and the passenger include: S21, define the utility function U of the government entity G =α1S cover -α2C cong , the utility function U of the enterprise entity E =β1R income -β2C oper , the utility function U of the passenger P =-γT total ; Among them, S cover is the service coverage, C cong is the congestion index, R income is operating income, C oper is the operating cost, T total To complete a total travel time, U G ,U E ,U P are the utility functions of the government, enterprise and passenger entities respectively; α1 and α2 are the weight coefficients of the influence of the preset service coverage and congestion index on the government utility; β1 and β2 are the weight coefficients of the influence of the preset operating income and operating cost on the enterprise utility; γ is the weight coefficient of the influence of the preset total time to complete a trip on the passenger utility.

4. The software planning decision-making support method based on spatiotemporal big data according to claim 3 is characterized in that: In step S2, the step of solving the game equilibrium solution by dynamically adjusting the bargaining power parameters includes: S22, based on real-time traffic event data, calculate the bargaining power parameters θ of the government, enterprises, and passengers respectively G (t),θ E (t),θ P (t), where: Passenger bargaining power parameter θ P (t) is calculated as follows: Among them, N event (t) is the number of traffic events in the current period, is the average number of traffic incidents in the same period of history, η P is the passenger bargaining sensitivity coefficient, The base value of passengers’ bargaining power; The government's bargaining power parameter θ G (t) is calculated as follows: in, Indicates the current connection coverage rate With target threshold The difference, η G is the government bargaining sensitivity coefficient, The base value of the government's bargaining power; The bargaining power parameter θ of the enterprise E (t) is calculated as follows: Among them, C oper is the enterprise operating cost, R income is the enterprise operating income, η E is the enterprise bargaining sensitivity coefficient, The base value of the enterprise's bargaining power; S23, the bottom-line utilities of the three parties, namely the government, enterprises and passengers, are Based on the utility function U of each party G ,U E ,U P And the bargaining power parameter θ calculated in real time G (t),θ E (t),θ P (t) Construct the following Nash negotiation objective function: in, They represent the utility gains of the government, enterprises, and passengers relative to their bottom-line utilities respectively; S24, solving the problem through an iterative optimization algorithm. When the utility value change rate of the three parties during the iteration process is less than the preset convergence threshold, the algorithm is judged to have converged, and the solution obtained at this time is output as the game equilibrium solution.

5. The software planning decision-making support method based on spatiotemporal big data according to claim 1 is characterized in that: In step S3, the step of using the game equilibrium solution as a constraint condition to construct a mixed integer collaborative optimization model includes: S31, defining decision variables of the mixed integer collaborative optimization model, wherein the decision variables include: Binary decision variable z j , when z j =1 means to choose to build connecting station j, when z j =0 means not to choose to build connecting station j; Positive integer decision variable f k , represents the departure frequency of planned bus route k; Continuous decision variable x ij , represents the proportion of travel demand of demand point i allocated to connecting station j; S32: Extract or determine the theoretical optimal allocation ratio of each demand point i to each potential connecting station j from the obtained game equilibrium solution, denoted as S33, constructing the objective function of the mixed integer collaborative optimization model to minimize the comprehensive cost, the comprehensive cost is the weighted total passenger travel time, the total operating cost of the enterprise, and the actual allocation ratio and the theoretical optimal allocation ratio The deviation between them is composed of the expression of the objective function: Among them, T total Represented by the decision variable z j ,f k ,x ij The total travel time of all passengers in the system is determined by C oper Represented by the decision variable z j ,f k ,x ij The total operating cost of the enterprise determined by ω1 and ω2 are the weight parameters of the total travel time of the passengers and the total operating cost of the enterprise respectively; λ is the preset actual allocation ratio x for the penalty ij Deviation from the theoretical optimal distribution ratio Deviation penalty coefficient; represents the actual allocation ratio x from all demand points i to all connecting stations j ij Compared with the theoretical optimal allocation ratio The total absolute deviation between S34, setting constraints for the mixed integer collaborative optimization model, wherein the constraints at least include: Demand satisfaction constraint: The allocation ratio of each demand point i satisfies ∑ j x ij =1; Connecting station capacity constraint: the total passenger flow of connecting station j does not exceed its capacity limit; Line departure frequency constraint: The departure frequency meets the preset range; Game equilibrium solution coordination constraint: the absolute value of the deviation between the demand allocation ratio and the equilibrium solution does not exceed the preset threshold.

6. The software planning decision-making support method based on spatiotemporal big data according to claim 5 is characterized in that: In step S3, the step of solving the connection station layout and departure frequency plan by using a two-stage algorithm includes: When screening candidate solutions in the first stage, based on the constraints of the game equilibrium solution coordination constraints, a set of candidate solutions that meet the conditions is generated through a greedy search algorithm; In the second stage of exact solution, the top M solutions with the smallest objective function value are selected from the generated candidate solution set as the initial solution, and the decision variable z in each initial solution is j and f k The mixed integer collaborative optimization model is iteratively optimized and solved by the branch and bound method until the final connection station layout and departure frequency plan is obtained, which can make the objective function reach the optimal state or meet the preset optimization termination conditions.

7. The software planning decision-making support method based on spatiotemporal big data according to claim 5 is characterized in that: In step S4, the step of generating a target weight adjustment signal based on the real-time passenger flow prediction error and dynamically correcting the objective function parameters in the mixed integer collaborative optimization model includes: S41, real-time monitoring of the passenger flow prediction error e of each connecting station j in the current period t j (t), the passenger flow prediction error e j (t) is calculated by the following formula: in, is the predicted passenger flow of connecting station j in time period t, is the actual observed passenger flow of connecting station j in time period t; S42, based on the real-time passenger flow prediction error e of all J connecting stations j (t), calculate the average passenger flow prediction error in time period t The calculation formula is: Where J is the total number of connection sites in the system; S43, using a PID controller, based on the calculated average passenger flow prediction error The passenger total travel time weight parameter ω1 and the enterprise operating cost weight parameter ω2 used in the objective function of the mixed integer collaborative optimization model are dynamically calculated and updated in the following manner: The updated value ω1(t) of the passenger total travel time weight parameter in time period t is calculated by the following formula: The updated value of the enterprise operating cost weight parameter in time period t, ω2(t), is calculated by the following formula: in, is the preset base value of the passenger's total travel time weight parameter; K p ,K i ,K d are the proportional, integral, and differential coefficients of the PID controller; is the cumulative integral value of the average passenger flow prediction error from the preset integration start time t0 to the current time period t, where e(τ) represents the average passenger flow prediction error at time τ; is the rate of change of the average passenger flow prediction error in the current period t; is the preset base value of the enterprise operating cost weight parameter; γ1 is the preset weight coupling coefficient; S44, the updated passenger total travel time weight parameter ω1(t) and the enterprise operating cost weight parameter ω2(t) are fed back in real time to the objective function of the mixed integer collaborative optimization model to replace the original corresponding weight parameters in the subsequent solution or iterative adjustment process of the model.

8. The software planning decision-making support method based on spatiotemporal big data according to claim 1 is characterized in that: The method further comprises the following steps: Provide a manual interface to adjust the base value parameters of the bargaining power of governments, businesses, and passengers.

9. A software planning decision support system based on spatiotemporal big data, used to execute the method according to any one of claims 1 to 8, characterized in that: include: Data fusion module, used to fuse multi-source spatiotemporal data to build a dynamically updated spatiotemporal knowledge graph; A game modeling module is used to establish an asymmetric Nash negotiation model among government entities, business entities, and passenger entities based on the dynamically updated spatiotemporal knowledge graph, and to solve the game equilibrium solution by dynamically adjusting bargaining power parameters; An optimization solution module uses the game equilibrium solution as a constraint to construct a mixed integer collaborative optimization model and solve the connection station layout and departure frequency plan through a two-stage algorithm; a feedback control module, generating a target weight adjustment signal based on the real-time passenger flow prediction error, and dynamically correcting the objective function parameters in the mixed integer collaborative optimization model; The decision output module outputs a visual decision plan including a heat map of the docking station layout, a dynamic curve of the departure frequency, and a comparison of the utility of multiple agents, and supports manual intervention.

10. The software planning decision support system based on spatiotemporal big data according to claim 9 is characterized in that: The optimization solution module includes: A candidate solution screening unit, used to screen the range of feasible solutions based on the game equilibrium solution constraints; A branch and bound solving unit is used to accurately solve integer variables within a range of candidate solutions.

Images