Shared parking resource matching method based on dynamic pricing and collaborative path planning

Abstract

The invention discloses a shared parking resource matching method based on dynamic pricing and collaborative path planning, and belongs to the field of intelligent traffic and sharing economy. The optimization method is provided for solving the problems that an existing platform is rigid in pricing, and matching and path planning are separated. The method comprises the following steps: 1, constructing a multi-factor dynamic pricing model fusing real-time supply and demand, location attributes and historical data, and introducing a price smoothing mechanism to balance market response and price stability; 2, establishing a collaborative optimization model with maximization of the total benefit of the system as a target, jointly deciding parking space distribution, pricing fine adjustment and path planning, and comprehensively planning and optimizing multiple targets such as a platform and a user; and thirdly, designing a two-stage hybrid heuristic algorithm, and ensuring the real-time solution quality in a complex scene by combining a graph theory and an improved genetic algorithm. The invention provides a thought for improving the parking resource utilization rate, the user satisfaction and the social traffic efficiency.

Description

Technical Field

[0001] This invention relates to the fields of intelligent transportation technology, sharing economy and operations optimization, and in particular to a system and method for integrated and collaborative optimization of dynamic pricing strategies, parking space resource matching and user driving route planning. Background Technology

[0002] With the acceleration of urbanization and the surge in motor vehicle ownership in my country, the problems of parking difficulties and illegal parking in urban core areas are becoming increasingly prominent, especially in large urban commercial centers, hospitals, and older residential areas. The imbalance between parking supply and demand has become a bottleneck restricting the improvement of urban traffic efficiency. At the same time, many parking lots built in conjunction with urban infrastructure projects experience significant vacancy during specific off-peak hours, resulting in extremely low resource utilization. Due to these existing problems, the shared parking model based on Internet of Things (IoT) technology has emerged, aiming to alleviate the supply-demand imbalance by opening up idle resources during off-peak hours. However, although existing shared parking platforms have achieved digital dissemination of parking space information to some extent, the following significant technical defects and application bottlenecks still exist in actual operation and resource scheduling, specifically manifested in the following ways:

[0003] 1. Lack of immediacy and predictability in pricing mechanisms. Most existing shared parking platforms adopt static pricing or simple time-based tiered pricing strategies. This model has significant lag. On the one hand, static prices cannot respond promptly to real-time market supply and demand fluctuations. For example, when demand surges due to sudden large-scale events or extreme weather, fixed low prices can lead to parking spaces being snapped up instantly, failing to identify high-value users in urgent need of parking and also failing to maximize revenue for parking space owners. On the other hand, existing dynamic pricing attempts are mostly based on the current supply-demand ratio, lacking the ability to predict short-term traffic flow and parking demand. This often results in a delayed price adjustment, failing to guide traffic flow through price levers in advance to smooth out peak and off-peak periods, and may even lead to frequent price fluctuations, reducing user trust.

[0004] 2. Resource matching and route planning are disconnected and lack synergy. In existing technical architectures, parking space allocation and route navigation are typically treated as two independent, sequential decision-making processes. Systems often prioritize recommending the nearest parking space to the user based on straight-line distance or static road network distance, only calling the map API for navigation after user confirmation. This sequential approach of matching first and then navigating ignores the decisive impact of dynamic traffic conditions on the decision. For example, during evening rush hour, the parking lot closest to the user may be located along severely congested roads, resulting in a significantly higher Actual Time of Arrival (ETA) and fuel / time costs compared to a slightly farther but less congested parking lot. Simply matching parking spaces while ignoring traffic conditions often leads to the system recommending a solution that does not minimize the user's overall cost.

[0005] 3. The optimization objective is singular, lacking a holistic consideration of social traffic efficiency. Existing algorithm models typically only take maximizing platform revenue or minimizing individual user distance as a single optimization objective. This locally optimal decision-making mechanism is prone to the fallacy of composition, that is, when the system simultaneously guides a large number of users to the same popular area or parking lot, it artificially creates local traffic congestion, thereby increasing the waiting time for all users and reducing the overall traffic efficiency of the regional road network.

[0006] Existing technologies lack a multi-objective dynamic weighting mechanism that can coordinate platform revenue, user experience, and social traffic congestion index. They cannot automatically adjust strategies to prioritize traffic flow when traffic congestion worsens, making it difficult to achieve a balance between social and economic benefits.

[0007] In summary, the key technical problem that urgently needs to be solved in the current field of shared parking technology is to build a system that can perceive changes in road conditions and supply and demand in real time, integrate and optimize dynamic pricing, parking space resource allocation and driving route planning, and improve social traffic efficiency while protecting the interests of all parties. Summary of the Invention

[0008] Purpose of the invention: To address the above problems, this invention provides a method for matching shared parking resources based on dynamic pricing and collaborative route planning, which enables efficient allocation of shared parking resources while improving platform revenue, user satisfaction, and social traffic efficiency.

[0009] Technical Solution: This invention discloses a method for matching shared parking resources based on dynamic pricing and collaborative route planning, comprising the following steps:

[0010] Step 1: Collect user demand data, parking space resource data, traffic network data, and parking space attribute feature vector data, and clean and standardize the data.

[0011] Step 2: Calculate the base price based on the parking space attribute feature vector, and dynamically adjust the base price by combining the real-time supply and demand ratio coefficient and time regularity factor generated in Step 1. Introduce a price smoothing mechanism and boundary truncation strategy to generate the execution price at the current moment.

[0012] Step 3: Establish a collaborative optimization model with the goal of maximizing the overall system benefits, which include platform revenue, user generalized costs, and social traffic congestion index; use a two-stage hybrid heuristic algorithm to jointly solve the parking space matching scheme and user driving path, and output the final matching result.

[0013] Furthermore, in step 1, the multi-source data includes user demand data. Parking space resource data Traffic network data Parking space attribute feature vector data .

[0014] Furthermore, in step 1, the user demand data in the multi-source data... Contains the user's current position vector Parking time window User price sensitivity coefficient .

[0015] Furthermore, in step 1, the parking space resource data in the multi-source data... Contains a real-time parking space status matrix Real-time supply-demand ratio coefficient .

[0016] Furthermore, in step 1, the traffic network data in the multi-source data... That is, a directed graph of a dynamic road network. .

[0017] Furthermore, in step 1, the multi-source data includes parking space attribute feature vector data. It converts the physical attributes of parking spaces into numerical values ​​that the algorithm can understand, and uses this to evaluate the quality of parking spaces.

[0018] Furthermore, in step 1, the user demand data in the multi-source data... Contains the user's current location vector This refers to obtaining latitude and longitude coordinates directly through the user's mobile phone GPS module, represented as... Convert WGS84 coordinates to the Mars coordinate system (GCJ-02) or Baidu coordinate system and match them with electronic maps.

[0019] Furthermore, in step 1, the user demand data in the multi-source data... Includes parking time window The estimated parking duration entered by the user in the app. and estimated arrival time The start time is The end time is ; recorded as This data is used to determine whether the parking space has been reserved or occupied.

[0020] Furthermore, in step 1, the user demand data in the multi-source data... Includes user price sensitivity coefficient It is a The values ​​within the range indicate that users are more concerned about price, while lower values ​​indicate that they are more concerned about distance or convenience; the calculation formula is... ,in, This represents the number of times a user canceled orders in the past 30 days due to excessively high prices. This represents the user's total number of orders placed in the past 30 days. The average historical transaction price per unit for users; This represents the average parking price per unit in the market for this region. These are the weighting coefficients; finally, the user... Price sensitivity coefficient This is used for personalized weight allocation in the subsequent step 3.

[0021] Furthermore, in step 1, the parking space resource data in the multi-source data... Contains a real-time parking space status matrix Describe what parking spaces are available, whether they are vacant, and their scarcity.

[0022] Furthermore, in step 1, the parking space resource data in the multi-source data... Contains a real-time parking space status matrix To report data for each parking space via the parking gate data interface or geomagnetic / video pile IoT device. Generate binary state variables If the parking space is occupied or has been reserved, then (A car is available or reserved); if the parking space is vacant and not reserved, then (Available and no reservations).

[0023] Furthermore, in step 1, the parking space resource data in the multi-source data... Includes real-time supply and demand ratio coefficient The core indicator reflecting the current shortage of parking resources in the region directly drives dynamic pricing; its calculation principle is: (1) set the search radius with the user's destination as the center. (2) Count the number of currently active search requests in this area. (3) Statistically determine the status of this area. Available parking spaces The calculation formula is: ,in, To minimize the value, prevent the denominator from being 0.

[0024] Furthermore, in step 1, the parking space resource data in the multi-source data... Includes real-time supply and demand ratio coefficient In order to prevent An excessively large value can cause prices to spiral out of control; the Sigmoid function is typically used to map it to... The range is used as a price multiplier factor and normalized.

[0025] Furthermore, in step 1, the traffic network data in the multi-source data... The system obtains real-time traffic information by calling the Gaode / Baidu Maps API, which is used to generate a weighted adjacency matrix for subsequent path planning algorithms, i.e., a dynamic directed road network graph. , where nodes It is a road intersection; side For road segments; edge weights are ,in, The physical length of the road segment This represents the current real-time average speed of vehicles on this road segment (considering congestion).

[0026] Furthermore, in step 1, the multi-source data includes parking space attribute feature vector data. Quantify static attributes, utilize Walking distance (in meters) from parking space to destination Parking space type (underground / ground / open-air) Parking difficulty (rated based on lane width and turning radius) is assessed by scoring walking distance and environment, with walking distance being the most important factor. The environment is divided into Safety is divided into The above scores are normalized and combined to form a parking space attribute score. .

[0027] Furthermore, step 2 aims to establish a multi-factor-linked pricing function. Calculate the first One parking space Final execution price at any time .

[0028] Furthermore, in step 2, the parking space attribute feature vector output in step 1 is used... Calculate the basic price of a parking space by combining the regional guidance price. Its calculation formula is ,in, The average parking price per unit in the market of this region (as a benchmark anchor point); Rate the attributes of the parking spaces; This refers to the weighting coefficient for parking space attributes. This means that the base price of a parking space is determined by the quality of its physical facilities; the higher the quality (such as proximity, underground parking, easy parking), the higher the base price.

[0029] Furthermore, in step 2, the basic price of the parking space... Dynamic corrections are made to supply and demand and time-varying factors, and the normalized real-time supply-demand ratio coefficient output from step 1 is introduced. And by adding a time regularity factor, the unsmoothed real-time theoretical price is calculated.

[0030] Furthermore, in step 2, a price elasticity adjustment coefficient is introduced. It is used to control the magnitude of price response to changes in supply and demand. The larger the value, the more drastic the price fluctuations caused by changes in supply and demand.

[0031] Furthermore, in step 2, a historical time adjustment factor is introduced. Based on the inherent popularity coefficient of the current moment (e.g., morning rush hour 8:00-9:00) obtained from historical big data mining; if no historical data is available, then... .

[0032] Furthermore, in step 2, the basic price of the parking space... The parking space price after dynamic adjustment for supply and demand and time-varying factors is Its calculation formula is , among which, when When supply falls short of demand, the price is... On the basis of rising; when When supply exceeds demand, prices decrease; It is used to correct the lag that may exist when relying solely on real-time data.

[0033] Furthermore, in step 2, a smoothing algorithm is introduced to calculate the smoothed price in response to the problem of frequent price fluctuations.

[0034] Furthermore, in step 2, a smoothing algorithm is introduced to calculate the smoothed price. ,in, Base price for parking spaces The price after dynamic adjustment based on supply and demand and time-varying factors; This represents the final price actually executed by the system at the previous moment; for ),in, When prices are relatively low, the price change curve is flat, and user acceptance is high; conversely, the market response is rapid when prices are high.

[0035] Furthermore, in step 2, in order to comply with regulatory requirements and prevent algorithmic anomalies, boundary truncation is required, introducing government-guided prices or minimum / maximum price limits set by the platform. To obtain the final parking space price Its calculation formula is .

[0036] Furthermore, in step 3, the multi-objective collaborative optimization modeling and hybrid heuristic solution aim to solve an NP-hard combinatorial optimization problem: under the premise of satisfying time and space constraints, for multiple users... Match the best parking space And plan the optimal path .

[0037] Furthermore, in step 3, during the multi-objective collaborative optimization modeling, variables are defined, including the current position vector. Extract into user set Real-time parking space status matrix Extract as a candidate parking space set Filter from the status in step 1 Parking spaces; road sections are grouped as That is, the set of directed road segments in the road network. A subset of.

[0038] Furthermore, in step 3, during multi-objective collaborative optimization modeling, the output (solution) of this model is determined by paired decision variables. and path decision change It is determined by two variables.

[0039] Furthermore, in step 3, in the multi-objective collaborative optimization modeling, paired decision variables... for:

[0040] .

[0041] Furthermore, in step 3, during multi-objective collaborative optimization modeling, path decision variables... for:

[0042] .

[0043] Furthermore, in step 3, the multi-objective collaborative optimization modeling includes three sub-objectives, which are weighted by coefficients. Adjustments are made, and the calculation formula is as follows: .

[0044] Furthermore, in step 3, the platform revenue maximization sub-model in the multi-objective collaborative optimization modeling... Its calculation formula is ,in, The parking spaces calculated in step 2 The final price.

[0045] Furthermore, in step 3, the user generalized cost minimization sub-model in the multi-objective collaborative optimization model. The calculation formula is as follows:

[0046] ,in, For the user in step 1 Price sensitivity; For the parking space in step 1 Walking distance score; The time it takes for a user to travel to a parking space, i.e. ( (The segment weights calculated in step 1). This is a time value coefficient, intended to convert minutes into monetary units; Penalty weights for walking experience.

[0047] Furthermore, in step 3, the sub-model for minimizing the social traffic congestion index is included in the multi-objective collaborative optimization modeling. Its calculation formula is ,in, For road section Basic traffic; The value is the exponent of the road resistance function (usually taken as 4.0). The larger the flow rate, the more exponentially the penalty increases.

[0048] Furthermore, in step 3, the constraints in the multi-objective collaborative optimization modeling are unique constraints.

[0049] Furthermore, in step 3, the uniqueness constraint in the multi-objective collaborative optimization modeling is: (1) (Each user must be matched with one and only one parking space); (2) (Each parking space can be allocated to a maximum of one user).

[0050] Furthermore, in step 3, the two-stage hybrid heuristic solution algorithm is divided into two stages: candidate set pruning based on spatiotemporal neighborhood and global optimization using an improved genetic algorithm (IGA).

[0051] Furthermore, in step 3, the candidate set pruning stage based on spatiotemporal neighborhood in the two-stage hybrid heuristic solution algorithm utilizes graph theory knowledge to eliminate physically or temporally impossible matching combinations before entering complex optimization, thereby changing the decision matrix from dense to sparse and significantly reducing the computational dimension.

[0052] Furthermore, in step 3, the candidate set pruning stage based on spatiotemporal neighborhood in the two-stage hybrid heuristic solution algorithm first prunes the spatial domain for each user. Iterate through all parking spaces Calculate straight-line distance or road network baseline distance, and set the maximum search radius. , build users Valid candidate set Its calculation formula is ,in, From step 1, From Step 1 Real-time Parking Space Status Matrix .

[0053] Furthermore, in step 3, the candidate set pruning stage based on spatiotemporal neighborhood in the two-stage hybrid heuristic solution algorithm further filters the time domain, utilizing the time window data from step 1. Eliminate infeasible solutions if the user expects to arrive at the destination. Parking space currently occupied until the end time, or estimated departure time. The parking space will be closed at the designated time. from Removed from the middle.

[0054] Furthermore, in step 3, the candidate set pruning stage based on spatiotemporal neighborhood in the two-stage hybrid heuristic solution algorithm finally pre-calculates the path cost for all retained legal pairs. (in Using Dijkstra's algorithm or A* algorithm on the road network map (with weights equal to those in step 1) Pre-calculate the shortest travel time. The significance of this step is that in subsequent iterations of the genetic algorithm, there is no need to repeatedly call the time-consuming pathfinding algorithm; the objective function can be calculated directly by looking up the table.

[0055] Furthermore, in step 3, the improved genetic algorithm (IGA) global optimization stage in the two-stage hybrid heuristic solution algorithm searches for a solution within the finite solution space generated in the first stage that makes the solution optimal. Maximize the globally optimal solution.

[0056] Furthermore, in step 3, the improved genetic algorithm (IGA) global optimization stage in the two-stage hybrid heuristic solution algorithm first performs chromosome encoding. In order to satisfy the constraint of "one user corresponds to one parking space", integer permutation encoding is used to define an individual (chromosome). For a length of The vector is calculated using the following formula: Among them, gene loci The value represents the value assigned to the first Each user's parking space ID (and must meet the following requirements) ),like This means that the user has not been matched with a parking space.

[0057] Furthermore, in step 3, during the global optimization phase of the improved genetic algorithm (IGA) in the two-stage hybrid heuristic solution algorithm, population initialization is performed again. To accelerate convergence, completely random generation is abandoned, and a hybrid strategy is used to generate the initial population. Among them, (1) 50% of greedy individuals: based on the nearest principle or The minimum principle directly generates local optimal solutions; (2) 50% random individuals: in Random selection is used to ensure population diversity.

[0058] Furthermore, in step 3, during the global optimization stage of the improved genetic algorithm (IGA) in the two-stage hybrid heuristic solution algorithm, the fitness function is selected, and the overall system benefit function is directly used as the fitness evaluation criterion. Its calculation formula is as follows: In this step, the pre-calculated data from the first stage is directly called. Substitution It has an extremely fast calculation speed.

[0059] Furthermore, in step 3, in the global optimization stage of the improved genetic algorithm (IGA) in the two-stage hybrid heuristic solution algorithm, an improved genetic operator is defined: (1) in the selection stage, an elite retention strategy is adopted, directly copying the 5% of individuals with the highest fitness in the previous generation to the next generation, and the rest are selected by roulette wheel selection; (2) in the adaptive crossover stage, the crossover probability is set. The population is dynamically adjusted based on its homogeneity, and a two-point crossover method is used to exchange partial gene fragments between two individuals. The calculation formula is as follows: (3) In the mutation process, for individuals A certain gene locus With probability Re-from the candidate set Randomly select another parking space replace.

[0060] Furthermore, in step 3, during the global optimization stage of the improved genetic algorithm (IGA) in the two-stage hybrid heuristic solution algorithm, crossover and mutation may lead to multiple users competing for the same parking space (i.e., and An illegal solution must be performed, and the steps are as follows: (1) Detect conflicting parking spaces (2) Obtain conflicting users and Price sensitivity coefficient (from step 1) (3) If (Right now If they are more price-sensitive and have a low willingness to pay, then they are judged as follows: Win, retain ;loser Forced reallocation: in its candidate set Find the next best available parking space; if no space is available, then... (Entering the waiting queue).

[0061] Furthermore, in step 3, the global optimization stage of the improved genetic algorithm (IGA) in the two-stage hybrid heuristic solution algorithm is set with a termination condition. The algorithm stops when any of the following conditions are met: (1) the number of iterations reaches... (2) Continuous generation, adaptability The increase is less than the threshold. .

[0062] Furthermore, in step 3, after the algorithm converges, it outputs the individual with the highest fitness. Decode it into decision variables and path decision change This is the final matching scheme.

[0063] The beneficial effects of this invention are: 1. This invention constructs a dynamic pricing model that is responsive and stable by integrating parking space attributes, real-time supply-demand ratio, time-based factors, and price elasticity adjustment coefficients; it introduces a price smoothing mechanism and boundary cutoff strategy to effectively avoid drastic price fluctuations, improve user price acceptance, and enhance the platform's market control capabilities; compared to traditional static or simple time-based pricing, this invention can more accurately reflect market supply and demand, guide parking resources towards high-value users, and thus significantly improve platform revenue and resource utilization.

[0064] 2. This invention integrates parking space allocation, pricing fine-tuning, and driving route planning into a single optimization framework, establishing a collaborative optimization model aimed at maximizing the overall system benefits. This model comprehensively considers platform revenue, user generalized costs including parking fees, travel time, walking distance, and the social traffic congestion index. It provides a new approach to the existing "matching first, then navigation" decision-making model. Through joint solution, the system can recommend parking spaces with the lowest overall cost to users and plan the optimal route based on real-time traffic conditions, improving user experience and travel efficiency.

[0065] 3. By introducing the social traffic congestion index as one of the optimization targets, this invention can proactively consider the impact of road network traffic flow in the matching decision, avoid local congestion caused by concentrating vehicles to a certain area, and automatically adjust the matching strategy during peak hours or congested road sections to disperse traffic flow, thereby reducing the overall traffic resistance of the regional road network, improving the efficiency of urban traffic operation, and achieving a balance between economic and social benefits.

[0066] 4. To address the large-scale, highly dynamic parking matching problem, this invention designs a two-stage hybrid heuristic algorithm. The first stage significantly compresses the solution space through spatiotemporal neighborhood pruning, while the second stage employs an improved genetic algorithm for global optimization and embeds a conflict resolution mechanism to ensure rapid output of high-quality feasible solutions even in complex scenarios. The algorithm supports real-time data input and online updates, meeting the requirements of shared parking platforms for response speed and decision-making accuracy.

[0067] 5. By introducing user price sensitivity coefficients and parking space attribute feature vectors, the system can implement differentiated matching and pricing strategies. Price-sensitive users are prioritized for recommending lower-cost parking spaces, while convenience-sensitive users are prioritized for recommending parking spaces that are close to their location and have good facilities, which helps improve user satisfaction and platform stickiness. Attached Figure Description

[0068] Figure 1 A flowchart of a shared parking resource matching method based on dynamic pricing and collaborative route planning.

[0069] Figure 2 Dynamic pricing model logic diagram overall flowchart

[0070] Figure 3 Flowchart of the candidate set pruning stage based on spatiotemporal neighborhood in a two-stage hybrid heuristic algorithm

[0071] Figure 4 Flowchart of global optimization in the improved genetic algorithm (IGA) within a two-stage hybrid heuristic algorithm Specific implementation examples

[0072] To better understand this invention, the following description, in conjunction with the accompanying drawings of the embodiments of this invention, will provide further details, but these descriptions are not intended to limit the scope of the invention. Various modifications and improvements made to the technical solutions of this invention by those skilled in the art without departing from the inventive concept should fall within the protection scope of this invention.

[0073] This embodiment uses Taishan District People's Hospital in Tai'an City as the application scenario, selecting 30 parking spaces in the hospital's underground parking lot (P1-P20), surface parking lot (P21-P28), and open-air temporary parking lot (P29-P30) as shared parking resources to serve 10 randomly selected patients (U1-U10). The implementation area is centered on the outpatient building of Taishan District People's Hospital, with a search radius of 1 kilometer, covering core surrounding roads such as Caiyuan Street, Qingnian Road, Dongyue Street, and Longtan Road. The method of this invention achieves coordinated optimization of parking space matching, dynamic pricing, and route planning.

[0074] like Figure 1As shown, a method for matching shared parking resources based on dynamic pricing and collaborative route planning includes the following steps:

[0075] Step 1: Collect user demand data, parking space resource data, traffic network data, and parking space attribute feature vector data, and clean and standardize the data.

[0076] Step 2: Calculate the base price based on the parking space attribute feature vector, and dynamically adjust the base price by combining the real-time supply and demand ratio coefficient and time regularity factor generated in Step 1. Introduce a price smoothing mechanism and boundary truncation strategy to generate the execution price at the current moment.

[0077] Step 3: Establish a collaborative optimization model with the goal of maximizing the overall system benefits, which include platform revenue, user generalized costs, and social traffic congestion index; use a two-stage hybrid heuristic algorithm to jointly solve the parking space matching scheme and user driving path, and output the final matching result.

[0078] Furthermore, in step 1, the multi-source data includes user demand data. Parking space resource data Traffic network data Parking space attribute feature vector data Please see Table 1 for details.

[0079] Table 1. Scope and Source of Data Collection

[0080]

[0081] Furthermore, in step 1, the user demand data in the multi-source data... Contains the user's current position vector Parking time window User price sensitivity coefficient See Table 2 for details.

[0082] Table 2.10 User Needs Data (Standardized)

[0083]

[0084] Note: Price sensitivity coefficient ∈ [0,1]. The higher the value, the more concerned the price is; the lower the value, the more concerned the distance / convenience is.

[0085] Furthermore, in step 1, the parking space resource data in the multi-source data... Contains a real-time parking space status matrix Describe what parking spaces are available, whether they are vacant, and their scarcity.

[0086] Furthermore, in step 1, the parking space resource data in the multi-source data... Includes real-time supply and demand ratio coefficient The core indicator reflecting the current shortage of parking resources in the region directly drives dynamic pricing; its calculation principle is: (1) set the search radius with the user's destination as the center. (2) Count the number of currently active search requests in this area. (3) Statistically determine the status of this area. Available parking spaces The calculation formula is: ,in, To minimize the value, prevent the denominator from being 0.

[0087] Furthermore, in step 1, the parking space resource data in the multi-source data... Includes real-time supply and demand ratio coefficient In order to prevent An excessively large value can cause prices to spiral out of control; the Sigmoid function is typically used to map it to... The range is used as a price multiplier factor and normalized. See Table 3 for details.

[0088] Table 3. Real-time Parking Space Status and Supply-Demand Ratio Calculation Table

[0089]

[0090] Furthermore, in step 1, the traffic network data in the multi-source data... The system obtains real-time traffic information by calling the Gaode / Baidu Maps API, which is used to generate a weighted adjacency matrix for subsequent path planning algorithms, i.e., a dynamic directed road network graph. , where nodes It is a road intersection; side For road segments; edge weights are ,in, The physical length of the road segment The current real-time average vehicle speed for this road segment (considering congestion) is shown in Table 4 for details.

[0091] Table 4. Regional Transportation Network Data Table

[0092]

[0093] Furthermore, in step 1, the multi-source data includes parking space attribute feature vector data. Quantify static attributes, utilize Walking distance (in meters) from parking space to destination Parking space type (underground / ground / open-air) Parking difficulty (rated based on lane width and turning radius) is assessed by scoring walking distance and environment, with walking distance being the most important factor. The environment is divided into Safety is divided into The above scores are normalized and combined to form a parking space attribute score. See Table 5 for details.

[0094] Table 5. Feature Vector Table of 30 Parking Space Attributes (Standardized)

[0095]

[0096] Note: Weighting: Walking distance 0.4, Parking space type 0.2, Parking difficulty 0.1, Safety score 0.2, Environment score 0.1; Type score: Underground = 1.0, Ground = 0.6, Open = 0.4; Parking difficulty score = (6 - Level) / 5.

[0097] Furthermore, step 2 aims to establish a multi-factor-linked pricing function. Calculate the first One parking space Final execution price at any time See Table 6 for details on the parameter settings.

[0098] Table 6. Pricing Parameter Settings

[0099]

[0100] Furthermore, in step 2, the parking space attribute feature vector output in step 1 is used... Calculate the basic price of a parking space by combining the regional guidance price. Its calculation formula is .

[0101] Furthermore, in step 2, the basic price of the parking space... Dynamic corrections are made to supply and demand and time-varying factors, and the normalized real-time supply-demand ratio coefficient output from step 1 is introduced. And by adding a time regularity factor, the unsmoothed real-time theoretical price is calculated.

[0102] Furthermore, in step 2, a price elasticity adjustment coefficient is introduced. It is used to control the magnitude of price response to changes in supply and demand. The larger the value, the more drastic the price fluctuations caused by changes in supply and demand.

[0103] Furthermore, in step 2, a historical time adjustment factor is introduced. Based on the inherent popularity coefficient of the current moment (e.g., morning rush hour 8:00-9:00) obtained from historical big data mining; if no historical data is available, then... .

[0104] Furthermore, in step 2, the basic price of the parking space... The parking space price after dynamic adjustment for supply and demand and time-varying factors is Its calculation formula is .

[0105] Furthermore, in step 2, a smoothing algorithm is introduced to calculate the smoothed price. ...

[0106] Furthermore, in step 2, in order to comply with regulatory requirements and prevent algorithmic anomalies, boundary truncation is required, introducing government-guided prices or minimum / maximum price limits set by the platform. To obtain the final parking space price Its calculation formula is The specific calculation results are shown in Table 7.

[0107]

[0108] Table 7. Detailed List of Dynamic Pricing for 30 Parking Spaces

[0109] Note: Parking spaces with S=0 are in an occupied / reserved state. The final execution price is only pre-calculated by the system and is not included in the current matching.

[0110] Furthermore, in step 3, during the multi-objective collaborative optimization modeling, variables are defined, including the current position vector. Extract into user set Real-time parking space status matrix Extract as a candidate parking space set Filter from the status in step 1 Parking spaces; road sections are grouped as That is, the set of directed road segments in the road network. A subset of.

[0111] Furthermore, in step 3, during multi-objective collaborative optimization modeling, the output (solution) of this model is determined by paired decision variables. and path decision change It is determined by two variables.

[0112] Furthermore, in step 3, the multi-objective collaborative optimization modeling includes three sub-objectives, which are weighted by coefficients. Adjustments are made, and the calculation formula is as follows: For specific parameter settings, please refer to Table 8.

[0113] Table 8. Parameter settings for the collaborative optimization model

[0114]

[0115] Furthermore, in step 3, the platform revenue maximization sub-model in the multi-objective collaborative optimization modeling... Its calculation formula is .

[0116] Furthermore, in step 3, the user generalized cost minimization sub-model in the multi-objective collaborative optimization model. The calculation formula is as follows:

[0117] .

[0118] Furthermore, in step 3, the sub-model for minimizing the social traffic congestion index is included in the multi-objective collaborative optimization modeling. Its calculation formula is .

[0119] Furthermore, in step 3, the uniqueness constraint in the multi-objective collaborative optimization modeling is: (1) (Each user must be matched with one and only one parking space) (2) (Each parking space can be allocated to a maximum of one user).

[0120] Furthermore, in step 3, the hard constraint condition of the time window in the multi-objective collaborative optimization modeling is: This means that the user's expected departure time must be earlier than the parking space's closing time. .

[0121] Furthermore, in step 3, the two-stage hybrid heuristic solution algorithm is divided into two stages: candidate set pruning based on spatiotemporal neighborhood and global optimization using an improved genetic algorithm (IGA).

[0122] Furthermore, in step 3, the candidate set pruning stage based on spatiotemporal neighborhood in the two-stage hybrid heuristic solution algorithm utilizes graph theory knowledge to eliminate physically or temporally impossible matching combinations before entering complex optimization, thereby changing the decision matrix from dense to sparse and significantly reducing the computational dimension.

[0123] Furthermore, in step 3, the candidate set pruning stage based on spatiotemporal neighborhood in the two-stage hybrid heuristic solution algorithm first prunes the spatial domain for each user. Iterate through all parking spaces Calculate straight-line distance or road network baseline distance, and set the maximum search radius. , build users Valid candidate set Its calculation formula is ,in, From step 1, From Step 1 Real-time Parking Space Status Matrix See Table 9 for details.

[0124] Table 9. Spatial Pruning

[0125]

[0126] Furthermore, in step 3, the candidate set pruning stage based on spatiotemporal neighborhood in the two-stage hybrid heuristic solution algorithm further filters the time domain, utilizing the time window data from step 1. Eliminate infeasible solutions if the user expects to arrive at the destination. Parking space currently occupied until the end time, or estimated departure time. The parking space will be closed at the designated time. from Removed from the list, see Table 10 for details.

[0127] Table 10. Time-based pruning

[0128]

[0129] Furthermore, in step 3, the candidate set pruning stage based on spatiotemporal neighborhood in the two-stage hybrid heuristic solution algorithm finally pre-calculates the path cost for all retained legal pairs. (in ), using Dijkstra's algorithm on the road network map (with weights equal to those in step 1) Pre-calculate the shortest travel time. See Table 11 for details.

[0130] Table 11. Detailed table of path pre-calculation (using Dijkstra's algorithm to find the shortest travel time)

[0131]

[0132]

[0133]

[0134] Note: All road segment weights are taken from the traffic network data table above. Travel time = sum of road segment weights × 60 (rounded to two decimal places); negative utility of walking = 0.005 yuan / meter × walking distance; time cost = 0.1 yuan / minute × travel time. All calculation results are rounded to two decimal places.

[0135] Furthermore, in step 3, the improved genetic algorithm (IGA) global optimization stage in the two-stage hybrid heuristic solution algorithm searches for a solution within the finite solution space generated in the first stage that makes the solution optimal. Maximize the globally optimal solution.

[0136] Furthermore, in step 3, the improved genetic algorithm (IGA) global optimization stage in the two-stage hybrid heuristic solution algorithm first performs chromosome encoding. In order to satisfy the constraint of "one user corresponds to one parking space", integer permutation encoding is used to define an individual (chromosome). For a length of The vectors are specifically encoded as shown in Table 12.

[0137] Table 12. Chromosome Coding Table

[0138]

[0139] Furthermore, in step 3, during the global optimization phase of the improved genetic algorithm (IGA) in the two-stage hybrid heuristic solution algorithm, population initialization is performed again (see Table 13 for the specific initialization table). To accelerate convergence, completely random generation is abandoned, and a hybrid strategy is used to generate the initial population. Among them, (1) 50% of greedy individuals: based on the nearest principle or The minimum principle directly generates local optimal solutions; (2) 50% random individuals: in Random selection is used to ensure population diversity.

[0140] Table 13. Detailed data table for population initialization

[0141]

[0142]

[0143] Furthermore, in step 3, during the global optimization stage of the improved genetic algorithm (IGA) in the two-stage hybrid heuristic solution algorithm, the fitness function is selected, and the overall system benefit function is directly used as the fitness evaluation criterion. Its calculation formula is as follows: .

[0144] Furthermore, in step 3, in the global optimization stage of the improved genetic algorithm (IGA) in the two-stage hybrid heuristic solution algorithm, an improved genetic operator is defined: (1) in the selection stage, an elite retention strategy is adopted, directly copying the 5% of individuals with the highest fitness in the previous generation to the next generation, and the rest are selected by roulette wheel selection; (2) in the adaptive crossover stage, the crossover probability is set. The population is dynamically adjusted based on its homogeneity, and a two-point crossover method is used to exchange partial gene fragments between two individuals. The calculation formula is as follows: (3) In the mutation process, for individuals A certain gene locus With probability Re-from the candidate set Randomly select another parking space replace.

[0145] Furthermore, in step 3, during the global optimization stage of the improved genetic algorithm (IGA) in the two-stage hybrid heuristic solution algorithm, crossover and mutation may lead to multiple users competing for the same parking space (i.e., and For any illegal solution, a repair operation must be performed. The specific repair values ​​are shown in Table 14.

[0146] Table 14 Complete Conflict Remediation Execution Data Table

[0147]

[0148]

[0149] Note: All conflict repairs strictly follow the invention rules. Priority determination is based on the price sensitivity coefficient (α). The lower the α value, the higher the priority. The reassigned parking spaces are all taken from the corresponding user's suboptimal candidate set (sorted in ascending order of generalized cost). When there are no suboptimal vacant parking spaces, the user enters the waiting queue (gene value is set to 0). After repair, the generalized cost and fitness value are dynamically adjusted according to the comprehensive benefits after parking space replacement, maintaining consistency with the data logic in the previous text. Furthermore, in step 3, the global optimization stage of the improved genetic algorithm (IGA) in the two-stage hybrid heuristic solution algorithm is set with a termination condition. The algorithm stops when any of the following conditions are met: (1) The number of iterations reaches (2) Continuous generation, adaptability The increase is less than the threshold. The specific iteration table is shown in Table 15 below.

[0150] Table 15. Iteration Process and Fitness Change

[0151]

[0152] Furthermore, in step 3, after the algorithm converges, it outputs the individual with the highest fitness. Decode it into decision variables This is the final matching scheme, as shown in Table 16.

[0153] Table 16. Final matching scheme and output results (based on the optimal solution after 100 iterations)

[0154]

[0155] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made in accordance with the spirit and essence of the present invention should be covered by the present invention.

Claims

1. A method for matching shared parking resources based on dynamic pricing and collaborative route planning, characterized in that, It includes the following steps: Step 1: Collect user demand data, parking space resource data, traffic network data, and parking space attribute feature vector data, and clean and standardize the data. Step 2: Calculate the base price based on the parking space attribute feature vector, dynamically adjust the base price by combining the real-time supply and demand ratio coefficient and time regularity factor, and introduce a price smoothing mechanism and boundary truncation strategy to generate the execution price at the current moment; Step 3: Establish a collaborative optimization model with the goal of maximizing the overall system benefits, which include platform revenue, user generalized costs, and social traffic congestion index; use a two-stage hybrid heuristic algorithm to jointly solve the parking space matching scheme and user driving path, and output the final matching result.

2. The method for matching shared parking resources based on dynamic pricing and collaborative path planning according to claim 1, characterized in that, In step 1, the user demand data includes the user's current location vector, parking time window, and user price sensitivity coefficient; the parking space resource data includes the real-time parking space status matrix and the real-time supply-demand ratio coefficient; the traffic network data is a dynamic directed graph of the road network containing real-time traffic conditions; and the parking space attribute feature vector data is used to quantify the physical attributes and environmental quality of the parking space.

3. The method for matching shared parking resources based on dynamic pricing and collaborative path planning according to claim 2, characterized in that, The user price sensitivity coefficient It is a The values ​​within the range indicate that users are more concerned about price, while lower values ​​indicate that they are more concerned about distance or convenience; the calculation formula is... ,in, This represents the number of times a user canceled orders in the past 30 days due to excessively high prices. This represents the user's total number of orders placed in the past 30 days. The average historical transaction price per unit for users; This represents the average parking price per unit in the market for this region. These are the weighting coefficients.

4. The method for matching shared parking resources based on dynamic pricing and collaborative path planning according to claim 2, characterized in that, In step 2, the specific calculation process of the dynamic pricing model is as follows: First, the parking space attribute score is calculated. Compared with the regional market average unit price Calculate the basic price of a parking space ; Secondly, a real-time supply-demand ratio coefficient is introduced. Time regularity factor and price elasticity adjustment coefficient Calculate the corrected real-time theoretical price Finally, a smoothing algorithm is used to calculate the smoothed price. The final execution price is obtained by truncating the price based on the set maximum and minimum price limits; where... As a smoothing factor, This represents the final price actually executed by the system at the previous moment.

5. The method for matching shared parking resources based on dynamic pricing and collaborative path planning according to claim 1, characterized in that, The objective function of the collaborative optimization model in step 3 is defined as follows: ,in, These are the weighting coefficients; For the platform revenue maximization sub-model, it represents the sum of prices for all successfully matched parking spaces; The user's generalized cost minimization sub-model covers parking fees, walking distance disutility, and driving time costs; This is a sub-model for minimizing the social traffic congestion index, which calculates the congestion penalty caused by traffic flow on road segments based on the road resistance function.

6. The method for matching shared parking resources based on dynamic pricing and collaborative route planning according to claim 1, characterized in that, The two-stage hybrid heuristic algorithm in step 3 includes: (1) The first stage is candidate set pruning based on spatiotemporal neighborhood, that is, for each user, the feasible parking space candidate set is filtered based on the maximum search radius; infeasible solutions with time conflicts are eliminated based on the user parking time window and the parking space opening time window; the shortest travel time of candidate matching pairs in the dynamic road network is pre-calculated using Dijkstra's algorithm or A* algorithm; (2) The second stage is global optimization of improved genetic algorithm, that is, in the solution space after pruning, the global optimal solution is found through chromosome encoding, population initialization, fitness evaluation and genetic operations.

7. The method for matching shared parking resources based on dynamic pricing and collaborative path planning according to claim 6, characterized in that, The improved genetic algorithm uses integer permutation encoding, where the value of the gene bit represents the parking space ID assigned to the corresponding user; the population initialization adopts a hybrid strategy, including locally optimal individuals generated based on greedy rules and randomly generated individuals; the fitness evaluation directly uses the objective function value of the collaborative optimization model.

8. The method for matching shared parking resources based on dynamic pricing and collaborative route planning according to claim 6, characterized in that, The improved genetic algorithm includes a conflict resolution mechanism: when crossover or mutation operations produce an illegal solution where multiple users compete for the same parking space, the price sensitivity coefficients of the conflicting users are compared; the user with the lower price sensitivity coefficient is determined to retain the parking space, while the user with the higher price sensitivity coefficient is forced to be reallocated, and an empty parking space is searched in their second-best candidate set. If no empty space is found, the user enters the waiting queue.

9. A method for matching shared parking resources based on dynamic pricing and collaborative route planning according to claim 1, characterized in that, In step 1, the real-time supply-demand ratio coefficient is calculated as follows: the search radius is set with the user's destination as the center, the number of active search requests and the number of available parking spaces in the area are counted, the ratio of the two is calculated and normalized by the Sigmoid function.

Images