A path planning method and system

Abstract

The application discloses a path planning method and system, which is applied to the technical field of path planning and comprises the following steps: acquiring traffic path data of a target area and position coordinate data of each path node in the target area; inputting the traffic path data and the position coordinate data into a pre-constructed path planning model to obtain an initial path planning scheme; determining initial weight priorities of the path nodes based on real-time state data of each path node; splitting the traffic path data based on the path nodes and determining resource consumption data of each traffic path; analyzing and processing acquired user information, adjusting the initial weight priorities based on a processing result to obtain various weight priorities; and correcting the initial path planning scheme to obtain a target path planning scheme. The path planning method and system effectively improve the reliability of path planning.

Description

Technical Field

[0001] This invention relates to the field of path planning technology, and in particular to a path planning method and system. Background Technology

[0002] Path planning technology can automatically solve for the optimal path from the origin to the destination, significantly improving mobility efficiency. As a core algorithmic foundation for fields such as urban traffic management and autonomous robot navigation, existing shortest path planning methods provide crucial support for the intelligent operation of these systems.

[0003] Existing shortest path planning technologies are ill-suited for scenarios such as scenic areas. Shortest path planning is based on the geographical coordinates of a set starting point, ending point, and waypoints. However, in scenarios such as scenic areas, the path planning schemes obtained by the shortest path planning methods are not conducive to users' reasonable and comfortable tours of the scenic area, which ultimately seriously restricts tourist satisfaction and the operational efficiency of the scenic area. Summary of the Invention

[0004] This invention provides a path planning method and system to solve the technical problem that path planning schemes obtained by the shortest path planning method cannot meet the needs of scenic spots and other locations, thereby improving the reliability of path planning schemes.

[0005] To address the aforementioned technical problems, this invention provides a path planning method and system, the method comprising: Acquire the travel path data of the target area and the location coordinate data of each path node in the target area; The travel path data and the coordinate data of each location are input into the pre-constructed path planning model to obtain the initial path planning scheme; Based on the real-time status data of each path node, the initial weight priority of each path node is determined. The travel path data is split based on each of the path nodes, and the resource consumption data of each travel path is determined. The acquired user information is analyzed and processed, and the initial weight priorities are adjusted based on the processing results to obtain the respective weight priorities; Using multi-objective optimization techniques and user constraint information as constraints, the initial path planning scheme is modified based on the resource consumption data and the weights and priorities of each resource to obtain the target path planning scheme.

[0006] Preferably, determining the initial weight priority of each path node based on the acquired real-time status data of each path node includes: The corresponding queuing time and the running data of the path node are extracted from each of the real-time status data. Based on each of the aforementioned queuing times, a first weighted priority sequence is obtained; Based on the aforementioned operational data, a second weight priority sequence is obtained; Based on the first weight priority sequence and the second weight priority sequence, the initial weight priority of each path node is obtained.

[0007] Preferably, the step of splitting the travel path data based on each of the path nodes and determining the resource consumption data for each travel path includes: The travel path data is split based on each of the path nodes to obtain the travel path between any two of the path nodes; For each of the aforementioned travel paths, the corresponding resource consumption data is obtained based on the determined travel method.

[0008] Preferably, the step of analyzing and processing the acquired user information, and adjusting the initial weight priority based on the processing results to obtain each weight priority includes: The acquired user information is processed to determine the value of interest for each path node; Based on the user information, the data on consumable resources is determined; The initial weight priority is adjusted based on the consumable resource data and each of the values ​​of interest to obtain the respective weight priority.

[0009] Preferably, the step of employing multi-objective optimization technology, using user constraint information as constraints, and revising the initial path planning scheme based on each of the resource consumption data and each of the weight priorities, to obtain the target path planning scheme, includes: Parse the user constraint information and construct the solution space; Using the initial path planning scheme as the initial population, the target path planning scheme is obtained by iterative evolution using a multi-objective optimization algorithm within the solution space. The evolution process is configured such that in each generation of evolution, a first optimization objective is obtained based on each of the resource consumption data, a second optimization objective is obtained based on each of the weight priorities, and operations are performed on the path nodes.

[0010] Another aspect of the present invention provides a path planning system, comprising: The acquisition module is used to acquire the travel path data of the target area and the position coordinate data of each path node in the target area; The initial scheme module is used to input the travel path data and the coordinate data of each location into a pre-constructed path planning model to obtain an initial path planning scheme; The initial priority module is used to determine the initial weight priority of each path node based on the real-time status data of each path node obtained. The splitting module is used to split the travel path data based on each of the path nodes and determine the resource consumption data of each travel path. The analysis module is used to analyze and process the acquired user information, and adjust the initial weight priority based on the processing results to obtain the weight priority. The correction module is used to employ multi-objective optimization techniques, taking user constraint information as constraints, and correcting the initial path planning scheme based on the resource consumption data and the weight priorities of each resource to obtain the target path planning scheme.

[0011] Preferably, the initial priority module includes: An extraction unit is used to extract the corresponding queuing time and the running data of the path node from each of the real-time status data. The first unit is used to obtain a first weighted priority sequence based on each of the queuing times; The second unit is used to obtain a second weight priority sequence based on each of the aforementioned operational data; A priority unit is used to obtain the initial weight priority of each of the path nodes based on the first weight priority sequence and the second weight priority sequence.

[0012] Preferably, the splitting module includes: A travel path unit is used to split the travel path data based on each of the path nodes to obtain the travel path between any two of the path nodes; The resource consumption data unit is used to obtain the corresponding resource consumption data for each travel path based on a determined travel method.

[0013] Preferably, the analysis module includes: The interest value unit is used to process the acquired user information and determine the interest value of each path node; A consumable resource unit is used to determine consumable resource data based on the user information. An adjustment unit is used to adjust the initial weight priority based on the consumable resource data and each of the values ​​of interest, to obtain each weight priority.

[0014] Preferably, the correction module includes: The parsing unit is used to parse the user constraint information and construct the solution space; An evolutionary unit is used to iteratively evolve the initial path planning scheme as the initial population within the solution space using a multi-objective optimization algorithm to obtain the target path planning scheme. The evolutionary process is configured to obtain a first optimization objective based on each of the resource consumption data and a second optimization objective based on each of the weight priorities in each generation of evolution, and to perform operations on the path nodes.

[0015] Compared with the prior art, the beneficial effects of the present invention are at least one of the following: This invention innovatively introduces a method based on real-time status data of each path node in a target area to determine its initial weight priority. This is achieved by acquiring and processing user information, and adjusting the initial weight priorities accordingly. This upgrades static geographical distance planning to dynamic priority planning that considers both real-time node status and personalized user needs. Furthermore, the invention breaks down the path data based on each path node, determines the resource consumption data for each path, and employs multi-objective optimization techniques. Using user constraints as conditions, the initial path planning is modified based on the resource consumption data and the weight priorities to obtain the target path planning. This achieves intelligent decision-making that automatically balances path resource consumption and high-priority node coverage while meeting specific user constraints. Breaking away from the limitations of traditional shortest path methods that only focus on physical distance, it can generate comfortable paths that not only conform to users' personalized execution habits but also take into account the real-time operating status of the region, significantly improving users' task execution satisfaction. At the same time, it improves the overall operational efficiency of the target area by optimizing resource allocation. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the path planning method in one embodiment of the present invention; Figure 2 This is a schematic diagram of the path planning system in one embodiment of the present invention; Figure label: Among them, 11. Acquisition module; 12. Initial scheme module; 13. Initial priority module; 14. Splitting module; 15. Analysis module; 16. Correction module. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0018] In the description of this invention, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0019] In the description of this invention, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to communication within two components. The terms "vertical," "horizontal," "left," "right," "upper," "lower," and similar expressions used herein are for illustrative purposes only and do not indicate or imply that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0020] In the description of this invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0021] Path planning technology can automatically solve for the optimal path from the starting point to the destination, significantly optimizing movement efficiency. As a core algorithmic foundation for fields such as urban traffic scheduling and autonomous robot navigation, existing shortest path planning methods provide crucial support for the intelligent operation of these systems. However, there are significant differences between scenic area path planning and urban traffic shortest path planning: urban traffic path planning aims to solve for the shortest physical path from the starting point to the destination, while scenic area path planning typically requires the starting and ending points to coincide (i.e., closed-loop tours), and the core objective is to optimize the tour experience.

[0022] This makes it difficult for existing shortest path planning methods to adapt to the complex needs of scenic area route planning. Specifically, existing technologies often use overly mechanical and linear methods for planning coordinates at various locations. Their core logic is usually based on greedy strategies or simple graph search algorithms—a simple "point-to-point" connection method. While this mathematically guarantees the shortest physical distance, it completely ignores the semantic attributes of the nodes themselves and the logical relationships between them. Directly applying this method often results in monotonous and boring routes. For example, tourists might be guided to remote support routes right after seeing the core attractions, or they might miss important sights along the way in order to shorten the distance by a few dozen meters, causing a chaotic tour rhythm, a fragmented experience, and ultimately severely restricting tourist satisfaction and the operational efficiency of the scenic area.

[0023] One embodiment of the present invention provides a path planning method applied to scenic areas. For details, please refer to [link / reference]. Figure 1 , Figure 1 The diagram shown is a flowchart of a path planning method according to one embodiment of the present invention, including: S1. Obtain the travel path data of the target area and the location coordinate data of each path node in the target area; S2. Input the travel path data and coordinate data of each location into the pre-built path planning model to obtain the initial path planning scheme; S3. Based on the real-time status data of each path node, determine the initial weight priority of each path node; S4. Based on each path node, the travel path data is split and the resource consumption data of each travel path is determined. S5. Analyze and process the acquired user information, and adjust the priority of each initial weight based on the processing results to obtain the priority of each weight. S6. Using multi-objective optimization technology, with user constraint information as the constraint condition, the initial path planning scheme is modified based on the resource consumption data and the weight priority of each resource to obtain the target path planning scheme.

[0024] First, acquire the access path data and the location coordinates of each path node within the target area. In a scenic area, obtaining this data provides accurate basic geographic support for route planning, ensuring that the planned routes conform to the actual layout of the scenic area and are feasible. This improves the accuracy and practicality of route planning, guarantees the visitor experience, and assists in scenic area management. Access path data refers to information related to all routes accessible to visitors within the scenic area, including the direction, length, and road conditions of main roads, walking trails, viewing corridors, and branch roads connecting various attractions. Path nodes refer to all key locations within the scenic area that support the tour route, specifically including entrances to core attractions, entrances and exits, rest areas, intersections, viewing platforms, and service facilities. Location coordinates refer to the precise geographic location information of these path nodes, typically expressed in latitude and longitude.

[0025] First, a drone-based LiDAR system is used to conduct a comprehensive scan of the entire scenic area, acquiring high-precision 3D point cloud data. Then, geographic information system (GIS) software is used to classify and process the collected point cloud data, separating various travel paths and extracting their vector information. Simultaneously, GPS positioning devices are deployed to collect on-site location data for each path node, recording the precise latitude and longitude coordinates of each node. Afterward, the collected path vector data and node coordinate data are compared and calibrated, outlier data is removed, and missing information is supplemented to ensure the completeness of the travel path data and the accuracy of the node coordinate data. Finally, a standardized dataset of scenic area travel paths and node coordinates is compiled, providing reliable support for subsequent path planning model input.

[0026] Preferably, the access route data and coordinate data of each location are input into a pre-built path planning model to obtain an initial path planning scheme. Inputting the access route data and coordinate data into the pre-built path planning model to obtain the initial path planning scheme is to leverage the algorithm's computing power to quickly generate tour routes that fit the scenic area layout. This improves planning efficiency and path rationality, providing basic tour guidance for tourists. The access route data consists of information related to routes available to tourists within the scenic area, including the direction, length, and road surface conditions of each route. The location coordinate data is the precise geographic latitude and longitude information of each path node. The path planning model is a computer program that automatically calculates the optimal or specified characteristic paths by inputting geographic data. The initial path planning scheme is a set of basic routes output by the model, without incorporating real-time conditions and user needs.

[0027] First, the preprocessed scenic area route vector data and node coordinate data are standardized in format, unifying the data projection coordinate system and field naming rules to generate a JSON format input file that the model can directly read. Then, the input file is deployed to the pre-built route planning model. This model uses an existing deep learning model as the core computing logic. Combined with constraints such as one-way traffic and speed limits in the scenic area, it first constructs a topological network containing node connection relationships and path attributes. Then, taking the scenic area entrance / exit or core attractions as the starting and ending points, it traverses and searches all feasible path combinations. At the same time, it selects a set of candidate paths with a path length less than a set threshold and covering the main attractions. Then, the candidate paths are initially scored based on two core indicators: the number of attractions covered by the path and the travel time. Finally, three differentiated initial route planning schemes are selected from the candidate paths. The schemes focus on the shortest tour time and the most attractions covered, as well as a balanced experience and low physical exertion, respectively, completing the generation and output of the initial schemes.

[0028] Furthermore, based on the real-time status data of each path node, the initial weight priority of each path node is determined. The corresponding queuing time and the running data of each path node are extracted from each real-time status data; a first weight priority sequence is obtained based on each queuing time; a second weight priority sequence is obtained based on each running data; and the initial weight priority of each path node is obtained based on the first and second weight priority sequences.

[0029] Determining initial weight priorities based on real-time status data of each path node is to quantify the impact of each node on traffic flow at the current moment, allowing path planning to dynamically adapt to on-site visitor flow and facility operation. This effectively avoids congested areas and faulty nodes, ensuring smooth tours and visitor safety. Real-time status data refers to monitoring information generated by each path node during its dynamic operation, including visitor flow, facility service status, and environmental traffic conditions. Queuing time is the cumulative time visitors wait at path nodes, representing the degree of congestion. Operational data includes the operational status information of the facilities supporting the path nodes, including whether the facilities are operating normally, their capacity, and service efficiency. The first weight priority sequence is a numerical sequence generated based on the queuing time of each node; the longer the queuing time, the higher the value and priority. The second weight priority sequence is a numerical sequence generated based on the quality of operational data for each node; the worse the facility's operational status, the higher the value and priority. The initial weight priority is a comprehensive impact level of the node determined by considering both queuing time and operational data, directly influencing the traffic selection strategy for nodes in path planning.

[0030] First, an IoT sensor network is deployed to collect real-time data at each path node in the scenic area. Infrared passenger flow counters are deployed at node entrances to obtain the number of people queuing and timing devices are used to calculate queuing time. Status monitoring sensors are installed at the rest areas, ticket gates, and service facilities along the nodes to collect operational data. Then, edge computing devices are used to clean and denoise the collected raw data, removing outliers and missing values ​​caused by sensor malfunctions. Simultaneously, the data format is standardized, uniformly converting it into a numerical format suitable for model calculation. Next, a first-weight priority sequence is constructed based on the cleaned queuing time data. A linear normalization method is used to map queuing time to a numerical range of 0 to 100, with higher values ​​corresponding to later queuing times. Furthermore, a time dimension is set... Dynamic weighting is used, with the weight of queuing time increasing to 70% during peak hours and decreasing to 50% during off-peak hours. A second weight priority sequence is then constructed based on operational data, assigning a binary value to whether the facility is operating normally (1 for normal operation and 10 for failure). The facility's capacity and service efficiency are normalized and mapped back to a numerical range of 0 to 100. Finally, the first and second weight priority sequences are weighted and merged, with the weighting coefficient dynamically adjusted according to the scenic area's peak and off-peak seasons and time periods. During peak seasons, queuing time is the primary weight (60%), and operational data is the secondary weight (40%). During off-peak seasons, the weights are adjusted to 40% and 60%, respectively. After merging, the initial weight priority for each path node is generated, completing the dynamic weight allocation process.

[0031] Furthermore, the travel path data is split based on each path node, and the resource consumption data for each travel path is determined. The travel path data is split based on each path node to obtain a travel path between any two path nodes; for each travel path, the corresponding resource consumption data is obtained based on the determined travel method.

[0032] Breaking down the travel path data based on each path node and determining the resource consumption data for each path is to refine the cost accounting of the scenic area's tour routes, providing data support for subsequent path optimization. The benefit is that it can accurately match tourist demand with the scenic area's resource carrying capacity, improving the rationality and practicality of path planning. Path nodes are key locations within the scenic area, such as entrances, exits, and rest areas for core attractions. Travel path data includes information on various routes available to tourists within the scenic area. Path breakdown involves decomposing the overall scenic area tour route into several short paths based on the connection relationships between path nodes. The travel path between any two path nodes refers to the specific route connecting these two key locations. Travel methods include two main types: walking and scenic area sightseeing vehicles. Resource consumption data represents various cost information incurred by tourists traveling a particular path, specifically time consumption, physical exertion, and energy consumption.

[0033] Piezoelectric pressure sensors and infrared passenger flow counters are embedded in the pedestrian walkway to capture instantaneous pedestrian flow and walking speed in real time. GPS and BeiDou dual-mode positioning modules and accelerometers are integrated into the onboard terminals of the sightseeing vehicles to record acceleration / deceleration frequency and average speed. Millimeter-wave radar is used to monitor vessel density and water flow speed in the waterways, while real-time rainfall and wind speed data are obtained through a meteorological bureau interface. The collected raw data is then input into an edge computing gateway for preprocessing. A Kalman filter algorithm is used to remove noise interference and unify timestamps. Combined with pre-stored slope elevation values, road surface friction coefficient tables, and ecological red line distribution maps from the scenic area's static database, a weighted summation formula is used to dynamically calculate the resource consumption data for each road segment. The time cost value is calculated by dividing the physical length of the road segment by... The system calculates the average travel speed after weather correction, plus the waiting time at ticket gates or narrow passages predicted by a queuing theory model. The physical exertion index is calculated by multiplying the sine of the road slope by the road surface roughness coefficient based on the metabolic equivalent standard, and then multiplying by the physical exertion decay factor derived from the analysis of users' historical exercise heart rate variability. The environmental carrying capacity pressure coefficient is calculated by dividing the current instantaneous pedestrian flow by the maximum safe carrying capacity of the road segment based on ecological vulnerability assessment, and then multiplying by the sensitivity weight coefficient of cultural relic protection areas or rare plant areas. Finally, a multi-dimensional vector data containing time cost, physical exertion index, and environmental carrying capacity pressure coefficient is generated, thereby accurately quantifying the actual travel cost of each road segment and ensuring that the planning system can avoid inferior routes that are short but steep and difficult to travel, overcrowded, or ecologically sensitive.

[0034] Next, the acquired user information is analyzed and processed. Based on the processing results, the initial weight priorities are adjusted to obtain the individual weight priorities. The acquired user information is processed to determine the value of interest for each path node; based on the user information, consumable resource data is determined; based on the consumable resource data and each value of interest, the initial weight priorities are adjusted to obtain the individual weight priorities.

[0035] The acquired user information is analyzed and processed. Based on the processing results, the initial weights and priorities are adjusted to obtain the final weight priorities. This ensures that the weight allocation of path nodes aligns with the needs of individual or group tourists, preventing path planning from deviating from actual tourist preferences. The benefit is improved tourist experience and path acceptance, making the planning scheme more targeted. User information refers to tourist-related data obtained by the scenic area through online reservation systems and offline registration, including age structure, travel preferences, travel methods, and physical condition. User information analysis and processing involves cleaning, organizing, and classifying the raw user data. The interest value of a path node measures the degree of tourist attention to that node; the closer the node's attraction type and service function match the tourist's preferences, the higher the interest value. Consumable resource data represents the upper limit of various costs that tourists can bear during their visit, including available time, physical capacity, and spending budget. The initial weight priorities are the basic influence levels determined based on the real-time status of the nodes. The adjusted weight priorities are the comprehensive influence levels of the nodes optimized based on tourist needs, directly determining the selection and ordering of nodes in the path planning.

[0036] First, collaborative filtering and content-based recommendation technologies are used to clean and integrate the acquired user historical travel trajectories, real-time physiological monitoring data, and explicit preference tags. This process determines the interest value for each path node, such as a specific viewing platform, cultural relic, or interactive experience area. The interest value is a personalized attractiveness score calculated based on the user's dwell time, photo frequency, and social media sharing popularity of similar attractions. Simultaneously, consumable resource data is determined based on the user's current physical condition, remaining travel time, and budget constraints. Specifically, consumable resource data refers to the user's maximum physical exertion threshold, available time window, and willingness to pay additional fees at the current moment. Then, an adaptive weight adjustment mechanism within a multi-objective optimization framework is employed. The fixed time cost coefficient, physical exertion coefficient, and interest matching coefficient in the initial weight priority are adjusted based on the interest value of each path node and the user's consumable resource data. The system performs nonlinear mapping and dynamic correction on the matching degree. For example, when it detects that a user's heart rate is too high, causing their expendable physical resources to fall below a preset threshold, the system automatically reduces the weight priority of physical exertion on steep sections and increases the weight priority of traffic efficiency on gentler routes. Or, when it identifies that a user has a very high interest in a certain type of cultural exhibit and has sufficient remaining time resources, it significantly increases the interest matching weight priority of the path leading to that type of exhibit. The implementation method is to construct a dual-tower neural network model containing user profile vectors and path feature vectors, output the real-time adjustment factor of each weight priority, and substitute it into the cost function of the heuristic search algorithm to recalculate the optimal solution. This can transform the uniform general navigation into personalized guidance for each user, ensuring that the generated route not only matches the user's current physical and mental endurance but also maximizes the satisfaction of their core tour interests, thereby optimizing the scenic spot tour experience under limited resource constraints.

[0037] Finally, a multi-objective optimization technique is employed, using user constraint information as the constraint condition, to revise the initial path planning scheme based on various resource consumption data and weight priorities, thereby obtaining the target path planning scheme. The user constraint information is analyzed to construct a solution space; using the initial path planning scheme as the initial population, the multi-objective optimization algorithm is used iteratively within the solution space to obtain the target path planning scheme. The evolutionary process is configured such that in each generation of evolution, a first optimization objective is obtained based on various resource consumption data, a second optimization objective is obtained based on various weight priorities, and operations are performed on the path nodes.

[0038] By employing multi-objective optimization techniques and using user constraint information as the constraint condition, the initial path planning scheme is modified based on various resource consumption data and weight priorities to obtain the target path planning scheme. This aims to balance the needs of scenic area operation and individual tourist preferences, and solve the problems of congestion and poor experience that may exist in the initial scheme. The advantage is that it can generate the optimal and realistic tour route, taking into account both tourist experience and efficient use of scenic area resources. Multi-objective optimization technology is an intelligent algorithm that simultaneously optimizes multiple mutually constraining objectives to find a balance among various needs. User constraints are the limitations that tourists cannot overcome during their visit, specifically the upper limit of available time, upper limit of physical exertion, and upper limit of visit budget. The solution space is the set of all feasible path planning schemes that satisfy the user constraints. The initial population is the initial input of the multi-objective optimization algorithm, i.e., the set of initial path planning schemes generated previously. The multi-objective optimization algorithm is an intelligent computing method that can handle multiple optimization objectives simultaneously. Iterative evolution is the process by which the algorithm continuously optimizes the path scheme through continuous iteration. The first optimization objective is the optimization direction determined based on resource consumption data, with the core being to reduce tourists' time, physical exertion, and energy consumption. The second optimization objective is the optimization direction determined based on weight priority, with the core being to prioritize path nodes with higher weights. The operation of path nodes involves adjusting the nodes contained in the path during the iteration process by retaining, adding, and deleting them. The target path planning scheme is the final visit route that satisfies all constraints and takes into account both optimization objectives after multiple rounds of optimization.

[0039] First, user constraints, such as the list of essential attractions to be visited, the longest acceptable hiking distance, and the budget limit, are analyzed to construct a solution space containing all feasible path combinations. Then, using the initial path planning scheme as the initial population, a non-dominated sorting genetic algorithm is used for iterative evolution within the solution space. The evolution process is configured such that in each generation, a first optimization objective (minimizing the overall travel cost) is obtained based on resource consumption data such as time cost, physical exertion index, and environmental carrying capacity coefficient. Simultaneously, a second optimization objective (maximizing the travel experience value) is obtained based on the weighted priorities of interest matching degree, tour comfort, and eco-friendliness. The system performs crossover operations on path nodes, connecting two... High-quality sections of different routes, such as scenic boardwalks and efficient sightseeing bus connections, are recombined, and mutation operations are performed to randomly replace some inefficient nodes or adjust the tour order to explore new possibilities. Then, a Pareto optimal solution set that satisfies both the user's hard constraints and performs well on both optimization objectives is selected through fast non-dominated sorting. Finally, a unique target path planning scheme is selected from the solution set based on the user's real-time preferences. This avoids the problem that traditional single-objective algorithms are prone to getting trapped in local optima, resulting in short routes with poor experiences or interesting routes with excessive physical exertion. It ensures that the generated routes achieve a perfect balance between tour efficiency and personalized experience while strictly adhering to the user's time and physical limitations.

[0040] Another embodiment of the present invention provides a path planning method applied to shopping malls, comprising: First, acquire the traffic path data and the location coordinates of each path node within the target shopping mall. In a shopping mall scenario, obtaining this data provides accurate basic geographic support for internal navigation and customer flow guidance, ensuring that the planned shopping paths conform to the mall's actual floor layout and are feasible. This improves the accuracy and practicality of path planning, guarantees the customer shopping experience, and assists in mall operation scheduling. Traffic path data refers to information related to all routes available to customers within the mall, including the direction, length, and floor material of main passageways, secondary passageways, escalators, elevator shafts, connecting corridors, and lobbies connecting various shops. Path nodes refer to all key locations within the mall that support the shopping path, specifically anchor store entrances, mall entrances / exits, atrium rest areas, elevator entrances, restrooms, service counters, and promotional booth locations. Location coordinate data refers to the precise indoor positioning information of these path nodes, typically represented by three-dimensional spatial coordinates. First, a handheld LiDAR combined with vision technology is used to perform a comprehensive scan of the entire shopping mall, acquiring high-precision 3D point cloud data. Then, geographic information system software is used to classify and process the collected point cloud data, separating various access paths and extracting their vector information. Simultaneously, Bluetooth Beacon positioning devices and ultra-wideband Bluetooth base stations are deployed to collect on-site location data for each path node, recording the precise 3D coordinates of each node. Afterward, the collected path vector data and node coordinate data are compared and calibrated, eliminating abnormal data caused by reflections from the glass curtain wall and supplementing missing information to ensure the integrity of the access path data and the accuracy of the node coordinate data. Finally, a standardized dataset of shopping mall access paths and node coordinates is compiled, providing reliable support for subsequent path planning model input.

[0041] Preferably, the access path data and coordinate data of each location are input into a pre-built path planning model to obtain an initial path planning scheme. Inputting the access path data and coordinate data into the pre-built path planning model to obtain the initial path planning scheme is to leverage the algorithm's computing power to quickly generate shopping routes that fit the mall layout. This improves planning efficiency and path rationality, providing customers with basic shopping guidance. The access path data refers to information related to routes available to customers within the mall, including the direction, length, and ground conditions of each route. The location coordinate data is the precise indoor three-dimensional coordinate information of each path node. The path planning model is a computer program that automatically calculates the optimal or specified characteristic paths by inputting geographic data. The initial path planning scheme is a set of basic routes output by the model, without incorporating real-time conditions and customer needs. First, the preprocessed shopping mall path vector data and node coordinate data are standardized in format, unifying the data projection coordinate system and field naming rules to generate a JSON input file that the model can directly read. Then, the input file is deployed to a pre-built path planning model. This model uses a deep learning model as its core computational logic, combining constraints such as one-way traffic and escalator speed limits in the shopping mall. It first constructs a topological network containing node connections and path attributes, then searches all feasible path combinations, starting and ending at mall entrances / exits or key anchor stores. Simultaneously, it selects candidate paths with path lengths less than a set threshold and covering major brand stores. Next, the candidate paths are initially scored based on two core indicators: the number of stores covered and travel time. Finally, three differentiated initial path planning schemes are selected from the candidate paths, focusing on the shortest shopping time, the most brand coverage, and a balanced experience with low physical exertion, respectively, thus generating and outputting the initial schemes.

[0042] Furthermore, based on the real-time status data of each path node, the initial weight priority of each path node is determined. The corresponding queuing time and operational data of each path node are extracted from each real-time status data point; a first weight priority sequence is obtained based on each queuing time; a second weight priority sequence is obtained based on each operational data point; and the initial weight priority of each path node is obtained based on the first and second weight priority sequences. Determining the initial weight priority based on the real-time status data of each path node is to quantify the impact of each node on traffic flow within the mall at the current moment, allowing path planning to dynamically adapt to the on-site customer flow and facility operation. The advantage is that it can effectively avoid congested areas and faulty nodes, ensuring smooth shopping and customer safety. Real-time status data consists of monitoring information generated by each path node within the mall during dynamic operation, including customer flow, facility service status, and environmental access conditions. Queuing time is the cumulative time customers spend waiting to pass through or enter a store at a path node, used to characterize the congestion level of the node. Operational data is the operational status information of the supporting facilities at the path nodes, including whether elevators are operating normally, restroom occupancy rates, and cashier service efficiency. The first weight priority sequence is a numerical sequence generated based on the queuing time of each node; the longer the queuing time, the higher the value and the higher the priority. The second weight priority sequence is a numerical sequence generated based on the quality of operational data for each node; the worse the facility's operational status, the higher the value and the higher the priority. The initial weight priority is a comprehensive impact level of the node determined by considering both queuing time and operational data, directly determining the pass-through strategy for nodes in path planning.First, an IoT sensor network is deployed to collect real-time data at various nodes along the mall's pathways. Infrared customer flow counters and video analytics cameras are deployed at node entrances to capture the number of people queuing and timing devices to calculate queuing time. Status monitoring sensors are installed at elevators, restrooms, and cash registers to collect operational data. Then, edge computing devices clean and denoise the raw data, removing outliers and missing values ​​caused by signal obstruction. Simultaneously, the data format is standardized, converting it into a numerical format suitable for model calculation. Next, a first-weight priority sequence is constructed based on the cleaned queuing time data. A linear normalization method is used to map queuing time to a numerical range of 0 to 100, with higher values ​​corresponding to later queuing times. Dynamic weighting is also set according to the time dimension. The weighting is as follows: during peak holiday periods, the weight of queuing time is increased to 70%, while during off-peak periods on weekdays it is reduced to 50%. A second weight priority sequence is then constructed based on operational data. A binary value is assigned to whether the elevator is operating normally, with 1 for normal operation and 10 for failure. The occupancy rate of restrooms and the service efficiency of the cashier are normalized and mapped to a numerical range of 0 to 100. Finally, the first and second weight priority sequences are weighted and merged. The weighting coefficients are dynamically adjusted according to the mall's peak and off-peak seasons and time periods. During promotional periods, queuing time is the main weight, accounting for 60%, and operational data is the secondary weight, accounting for 40%. During weekdays, the weights are adjusted to 40% and 60%, respectively. After merging, the initial weight priority of each path node is generated, completing the dynamic weight allocation process.

[0043] Furthermore, the travel path data is broken down based on each path node, and the resource consumption data for each path is determined. Breaking down the travel path data based on each path node yields the travel path between any two path nodes; for each travel path, the corresponding resource consumption data is obtained based on the determined travel method. This breakdown of the travel path data based on each path node and determination of the resource consumption data for each path is to refine the cost accounting of shopping mall routes, providing data support for subsequent route optimization. The benefit is that it can accurately match customer needs with the mall's resource capacity, improving the rationality and practicality of route planning. The path nodes are key locations within the mall, such as the entrances and exits of anchor stores and rest areas. The passage path data includes information on various routes available for customers to pass through the mall. Passage path decomposition breaks down the overall shopping route of the mall into several short paths according to the connection relationship of the path nodes. The passage path between any two path nodes refers to the specific route connecting these two key locations. The passage mode includes three main types: walking, taking escalators, and elevators. The resource consumption data is information on various costs incurred by customers when passing through a certain path, specifically time consumption, physical consumption, and energy consumption. Piezoelectric pressure sensors and Wi-Fi probes are embedded in the ground of main passageways to capture instantaneous pedestrian flow and speed in real time. Operational status monitoring modules are integrated into the terminals of escalators and elevators to record start / stop frequency and average carrying rate. Simultaneously, the system accesses the mall's internal environmental monitoring system to obtain real-time temperature, humidity, and air quality data. The collected raw data is then input into an edge computing gateway for preprocessing. A Kalman filter algorithm is used to remove noise interference and unify timestamps. This data is then combined with pre-stored floor elevation differences, ground friction coefficient tables, and promotional activity distribution maps from the mall's static database. A weighted summation formula is used to dynamically calculate the resource consumption data for each segment. The time cost is calculated by dividing the physical length of the segment by the average time cost after correction for pedestrian density. The average passage speed is combined with the elevator waiting time or popular store entry waiting time predicted based on queuing theory models. The physical exertion index is obtained by multiplying the vertical height difference between floors and the ground roughness coefficient according to the metabolic equivalent standard, and then multiplying it by the physical exertion decay factor derived from the analysis of customers' historical step frequency data. The energy consumption coefficient is obtained by dividing the current instantaneous traffic flow by the maximum safe capacity of the passage determined based on fire evacuation requirements, and multiplying it by the sensory interference weight coefficient of high noise area or strong light area. Finally, a multi-dimensional vector data containing time cost, physical exertion index and energy consumption coefficient is generated, so as to accurately quantify the actual passage cost of each section and ensure that the planning system can avoid inefficient paths that are short but crowded and noisy, or have elevator malfunctions or sensory discomfort.

[0044] Next, the acquired user information is analyzed and processed. Based on the processing results, the initial weight priorities are adjusted to obtain the individual weight priorities. The process of processing the acquired user information determines the interest value of each path node; based on the user information, consumable resource data is determined; and based on the consumable resource data and each interest value, the initial weight priorities are adjusted to obtain the individual weight priorities. This analysis and processing of the acquired user information, and the adjustment of the initial weight priorities based on the processing results, ensures that the weight allocation of path nodes aligns with the needs of individual or group customers, preventing path planning from deviating from actual customer preferences. The benefit is improved customer shopping experience and path acceptance, making the planning scheme more targeted. User information refers to customer data obtained by the shopping mall through its membership app system and offline parking registration, including age structure, consumption preferences, companions, and physical condition. User information analysis and processing involves cleaning, organizing, classifying, and mining the raw user data. The interest value of a path node measures the degree of customer attention to a particular path node. The more the brand type and service function of the node matches the customer's preferences, the higher the interest value. Consumable resource data represents the upper limit of various costs that customers can bear during the shopping process, including available time, physical capacity, and spending budget. The initial weight priority is the basic influence level determined based on the real-time status of the node. The adjusted weight priority is the comprehensive influence level of the node after optimization based on customer needs, which directly determines the selection and ordering of nodes in the path planning.First, collaborative filtering algorithms and content-based recommendation technology are used to clean and integrate the acquired customer historical shopping trajectories, real-time location signaling data, and explicit brand preference tags. This process determines the interest value for each path node, such as a specific brand counter, children's playground, or dining area. The interest value refers to a personalized attractiveness score calculated based on the customer's dwell time with similar brands, fitting frequency, and membership point redemption activity. Simultaneously, consumable resource data is determined based on the customer's current physical condition, remaining shopping time, and budget constraints. Specifically, consumable resource data refers to the customer's maximum physical exertion threshold, available time window, and willingness to pay additional service fees at the current moment. Then, an adaptive weight adjustment mechanism within a multi-objective optimization framework is employed. The fixed time cost coefficient, physical exertion coefficient, and interest matching coefficient in the initial weight priority are adjusted based on the interest value of each path node and the customer's consumable resource data. The matching degree is nonlinearly mapped and dynamically corrected. For example, when it is detected that a customer is carrying an infant and toddler, causing their physical energy resources to fall below a preset threshold, the system automatically reduces the weight priority of physical energy consumption on stairwells without elevators and increases the weight priority of passage efficiency on direct elevator routes. Or, when it is identified that a customer has a very high interest in a certain type of beauty brand and has sufficient remaining time resources, the interest matching weight priority of the route to the brand's counter is significantly increased. The method is to construct a dual-tower neural network model containing customer profile vectors and path feature vectors, output the real-time adjustment factors of each weight priority, and substitute them into the cost function of the heuristic search algorithm to recalculate the optimal solution. This can transform the uniform general wayfinding into personalized shopping guides, ensuring that the generated routes not only meet the customer's current physical and mental endurance but also maximize the satisfaction of their core shopping interests, thereby optimizing the shopping experience in the mall under limited resource constraints.

[0045] Finally, a multi-objective optimization technique is employed, using user constraints as conditions, to revise the initial path planning scheme based on resource consumption data and weight priorities, resulting in the target path planning scheme. The user constraints are analyzed to construct a solution space; using the initial path planning scheme as the initial population, the multi-objective optimization algorithm iteratively evolves within the solution space to obtain the target path planning scheme. The evolution process is configured such that in each generation, a first optimization objective is obtained based on resource consumption data, a second optimization objective is obtained based on weight priorities, and operations are performed on the path nodes. This multi-objective optimization technique, using user constraints as conditions, and revising the initial path planning scheme based on resource consumption data and weight priorities to obtain the target path planning scheme, aims to balance the needs of mall operations and individual customer preferences, addressing potential congestion and poor customer experience issues in the initial scheme. Its advantage is the generation of optimal and realistic shopping routes, balancing customer experience and efficient utilization of mall resources. Multi-objective optimization technology is an intelligent algorithm that simultaneously optimizes multiple mutually constraining objectives to find a balance among various needs. User constraints are the limitations that customers cannot overcome during shopping, specifically including the upper limit of available time, the upper limit of physical exertion, and the upper limit of spending budget. The solution space is the set of all feasible path planning schemes that satisfy the user constraints. The initial population is the initial input of the multi-objective optimization algorithm, i.e., the set of initial path planning schemes generated previously. The multi-objective optimization algorithm is an intelligent computing method that can handle multiple optimization objectives simultaneously. Iterative evolution is the process by which the algorithm continuously optimizes the path scheme through continuous iteration. The first optimization objective is the optimization direction determined based on resource consumption data, with the core being to reduce the time, physical exertion, and energy consumption of customers during shopping. The second optimization objective is the optimization direction determined based on weight priority, with the core being to prioritize path nodes with higher weights. The operation of path nodes is the adjustment action of retaining, adding, and deleting nodes contained in the path during the iteration process. The target path planning scheme is the final shopping route that satisfies all constraints and takes into account both optimization objectives after multiple rounds of optimization.First, user constraints, such as the list of core brands that must be purchased, the longest acceptable walking distance, and the budget limit, are analyzed to construct a solution space containing all feasible path combinations. Then, using the initial path planning scheme as the initial population, a non-dominated sorting genetic algorithm is used for iterative evolution within the solution space. The evolution process is configured such that in each generation, a first optimization objective (minimizing the overall travel cost) is obtained based on resource consumption data such as time cost, physical exertion index, and energy consumption coefficient. Simultaneously, a second optimization objective (maximizing the shopping experience value) is obtained based on the weighted priorities of interest matching degree, shopping comfort, and environmental friendliness. The system performs crossover operations on path nodes, connecting two paths... High-quality sections of different routes, such as scenic atrium corridors and efficient elevator connections, are reorganized and mutation operations are performed to randomly replace some inefficient nodes or adjust the order of shopping to explore new possibilities. Then, a Pareto optimal solution set that satisfies both the customer's hard constraints and performs well on both optimization objectives is selected through fast non-dominated sorting. Finally, a unique target path planning scheme is selected from the solution set based on the customer's real-time preferences. This avoids the problem that traditional single-objective algorithms are prone to getting trapped in local optima, resulting in short routes with poor experiences or interesting routes with excessive physical exertion. It ensures that the generated routes achieve a perfect balance between shopping efficiency and personalized experience while strictly adhering to the customer's time and physical limitations.

[0046] It is important to note that the effective execution of the aforementioned path planning method highly depends on the accurate collection of basic geographic data and the reliable transmission of real-time status information. If there are deviations in the travel path data or delays in node status perception, the weight priority and resource consumption data calculated above will not be reliably implemented. Therefore, this embodiment, based on the explanation of the core planning logic, further provides the physical foundation supporting the implementation of this planning method. Preferably, it provides a deployment scheme for a scenic area travel path data collection and node status monitoring system. Regarding the existing technical content involved in this system, such as the selection of conventional sensors, general data transmission protocols, and basic data cleaning methods, those skilled in the art can make corresponding decisions based on the actual deployment needs of the scenic area, and will not be specifically described in this embodiment of the invention. The following will focus on describing the scenic area path planning support methods that are strongly related to the above steps S1 to S6, including key deployment steps to ensure that real-time status data can be accurately converted into weight priority, such as dynamic positioning of path node sensors and data collection threshold calibration, thereby ensuring the integrity and feasibility of the overall technical solution.

[0047] Another embodiment of the present invention provides a scenic route recommendation method based on large language model visual understanding and graph aggregation technology, aiming to solve the technical problems in the prior art, such as insufficient expression of scenic spot features, low accuracy of user constraint matching, and lack of global relevance in route planning. Through the construction of a full-process architecture of "visual semantic completion - enhanced feature extraction - constraint-aware reasoning - reinforcement learning optimization", this method realizes the intelligent generation of personalized multimodal navigation solutions from the original map data.

[0048] In the data input and preprocessing stage, the embodiment of the present invention first receives electronic map or geographic information system data through the scenic area map information input module, extracts the geometric topological skeleton of the scenic area using image processing technology, identifies path connectivity and calculates physical distances, and establishes an initial graph structure that only contains position and connection relationships. Subsequently, the system uses the optical character recognition and visual understanding capabilities of the vision-language large model to perform high-dimensional semantic filling on the map nodes, automatically reads text annotations, identifies the meanings of icons, and analyzes terrain features, and aligns the structured text with the coordinate data. On this basis, the system converts multi-dimensional attributes such as the name, type, estimated visit duration, supporting facilities, physical exertion level, and adjacent relationships of each scenic spot into a standard Token sequence.

[0049] Token is the basic text processing unit in the field of computer natural language processing, and its Chinese interpretation is "lexical unit" or "token". When converting multi-dimensional attributes such as the name, type, and estimated visit duration of scenic spots into a standard Token sequence, the system first uses a word segmentation algorithm to cut the continuous original text string into individual smallest segments with independent semantics, and these segments are Tokens. The form of Token is very flexible. It can be a complete word (such as "ropeway"), a Chinese character (such as "mountain"), or even a punctuation mark or sub-word segment (such as splitting "sightseeing bus" into "sightseeing" and "bus"). The system maps each segmented Token into a high-dimensional digital vector through the word embedding matrix of the pre-trained model and arranges them in order to form a sequence. This ordered list composed of digital vectors is the Token sequence, which is the standardized mathematical expression form of unstructured text data that can be directly read, calculated, and understood by artificial intelligence models.

[0050] Specifically, the word embedding matrix of a pre-trained language model is used to map text information into high-dimensional vectors, and positional encoding is added to distinguish different elements in the sequence. Short sequences are padded, and excessive sequences are truncated. The [CLS] marker at the beginning of the sequence is used to represent the overall semantic features of the attraction. [CLS] is an abbreviation for Classification. In pre-trained language models (such as BERT) in natural language processing, it is a special marker, usually forced to be placed at the very beginning of the input token sequence. Its main function is to aggregate and represent the global semantic features of the entire input sequence (in this case, representing all the attribute information of a attraction, such as name, type, duration, etc.) after the model has undergone multi-layer neural network encoding. During training, the model specifically optimizes the vector at this position to capture the overall meaning of the context. Therefore, in subsequent tasks, the system directly extracts the output vector corresponding to the [CLS] marker, which can be used as the overall feature representation of the attraction for classification, similarity calculation, or as the initial embedding of nodes in a graph neural network, without needing to perform averaging or max pooling on all vectors in the entire sequence. Meanwhile, the system constructs an undirected weighted graph to describe the spatial structure of the scenic area. Nodes in the graph represent various attractions. If two attractions are directly connected on the physical map, an edge is established between them. The weight calculation of the edge no longer relies solely on physical distance, but is based on a comprehensive consideration of physical distance and the similarity of attraction attributes. Specifically, the shorter the physical distance between two attractions, and the more similar or complementary their type attributes, the higher the connection weight between them. The generation of this weight matrix implies prior knowledge of "route convenience" and "experience rationality" in the graph structure.

[0051] In the core feature extraction stage, this embodiment of the invention employs a scenic spot enhanced aggregation feature extraction module, which performs layered iterative fusion based on a graph neural network nested Transformer architecture. This mechanism includes multiple identical processing layers, each sequentially executing three steps: node feature extraction, graph neural network aggregation, and graph-enhanced token encoding. First, the system extracts [CLS] vectors representing the overall semantics of the scenic spot from the token sequence of the current layer, forming node-level embeddings. Next, a graph neural network is used to perform neighborhood aggregation on these node embeddings; this operation maps the node embeddings to query vectors, key vectors, and value vectors, and uses the edge weight matrix constructed in the previous step as an attention mask to perform a weighted summation of the information of neighboring nodes, thereby generating enhanced node embeddings that incorporate neighborhood contextual information. Subsequently, the system performs the crucial graph-enhanced token encoding operation, introducing an asymmetric attention mechanism to achieve deep coupling between text and graph structure. In this mechanism, the vector generated from the original text token sequence is used as the query vector to preserve the semantic focus of the scenic spot itself; while the enhanced node embeddings after graph neural network aggregation are expanded and inserted into the head of the token sequence as key and value vectors, providing rich contextual background information. Through this asymmetric attention computation, the model can proactively retrieve and absorb surrounding topological environment information, guided by the semantics of the attraction itself. The calculated attention output, after residual connection and layer normalization processing, serves as the input for the next layer iteration. This "extraction-aggregation-encoding" process is repeated across multiple layers, enabling attraction features to perceive topological relationships at greater distances within the deep network. Finally, in this embodiment of the invention, the vector corresponding to the [CLS] marker is extracted again from the token sequence output by the last layer iteration, serving as the final enhanced aggregated feature for the attraction. This feature vector is a low-dimensional dense vector that not only fully preserves the original textual semantics of the attraction but also deeply integrates distance constraints, type associations, and global location context within the entire scenic area's road network of surrounding attractions.

[0052] In the constraint-aware topology reasoning stage, this embodiment of the invention receives the aforementioned enhanced aggregated features of scenic spots and user-structured constraint parameters, achieving accurate matching through multi-dimensional calculations. For user preference constraints, the semantic space isomorphic mapping principle is used to transform user preference descriptions into feature vectors. A preference matching score is calculated based on cosine similarity, reflecting the degree of fit between the implicit attributes of scenic spots and user interests. For numerical constraints such as time and physical exertion, a soft constraint relaxation mechanism is designed. A lightweight decoder predicts attribute values ​​from scenic spot features. Full marks are awarded when the predicted value meets the constraints; scores decay exponentially with the amount of deviation when the constraints are exceeded, thus avoiding reward sparsity. For path constraints, state transition costs are calculated based on Markov connectivity evaluation, combined with adjacency matrices and physical distance, ensuring the physical feasibility of the route. Finally, a comprehensive matching score for a single scenic spot is obtained based on the weighted sum of the scores from each dimension, providing a quantitative basis for subsequent planning.

[0053] In the route generation and optimization phase, this embodiment of the invention employs a generalized reinforcement learning strategy optimization framework, constructing a ternary multi-objective reward function consisting of constraint fit, spatial rationality, and logical interpretability. The constraint fit reward calculates the average of the comprehensive matching scores of attractions along the route and introduces a time penalty coefficient to ensure the route both aligns with interests and strictly adheres to time constraints. The path rationality reward calculates the ratio of the actual path distance to the theoretical shortest distance and penalizes the number of backtracking attempts and unreachable paths to eliminate detours. The reasoning format reward ensures the interpretability of the results by detecting whether the output contains complete structured thought chain labels and logical consistency. The system adopts a progressive strategy of greedy hot-start and reinforcement learning strategy optimization. It first constructs a baseline route using a heuristic algorithm, then iteratively optimizes it through steps such as group sampling, multi-dimensional reward evaluation, advantage estimation, and policy gradient updates, ultimately outputting the optimal route with the highest reward value.

[0054] Finally, this embodiment of the invention transforms the abstract optimal strategy into a multimodal tour guide solution that users can understand through the optimal route output module. This module not only outputs the specific tour route but also outputs a natural language explanation containing the complete reasoning process, making the algorithm's black box transparent. This allows users to clearly understand the planning basis, thereby significantly improving their trust and satisfaction with the recommendation results.

[0055] Another embodiment of the present invention provides a path planning system; for details, please refer to [link to relevant documentation]. Figure 2 , Figure 2 The diagram shown illustrates the structure of a path planning system according to one embodiment of the present invention, comprising: The acquisition module 11 is used to acquire the travel path data of the target area and the location coordinate data of each path node in the target area; The initial scheme module 12 is used to input the travel path data and the coordinate data of each location into the pre-built path planning model to obtain the initial path planning scheme; The initial priority module 13 is used to determine the initial weight priority of each path node based on the real-time status data of each path node. The splitting module 14 is used to split the travel path data based on each path node and determine the resource consumption data of each travel path. Analysis module 15 is used to analyze and process the acquired user information, and adjust the priority of each initial weight based on the processing results to obtain the priority of each weight. The correction module 16 is used to use multi-objective optimization technology, with user constraint information as the constraint condition, to correct the initial path planning scheme based on various resource consumption data and various weight priorities, so as to obtain the target path planning scheme.

[0056] Preferably, the initial priority module 13 includes: The extraction unit is used to extract the corresponding queuing time and the running data of the path node from each real-time status data. The first unit is used to obtain the first weight priority sequence based on each queuing time; The second unit is used to obtain the second weight priority sequence based on each running data. The priority unit is used to obtain the initial weight priority of each path node based on the first weight priority sequence and the second weight priority sequence.

[0057] Preferably, the split module 14 includes: The passage path unit is used to split the passage path data based on each path node to obtain the passage path between any two path nodes; The resource consumption data unit is used to obtain the corresponding resource consumption data for each travel path based on a determined travel method.

[0058] Preferably, the analysis module 15 includes: The interest value unit is used to process the acquired user information and determine the interest value for each path node; Consumable resource unit, used to determine consumable resource data based on user information; The adjustment unit is used to adjust the initial weight priority based on the consumable resource data and each value of interest to obtain the weight priority.

[0059] Preferably, the correction module 16 includes: The parsing unit is used to parse user constraint information and construct the solution space; An evolutionary unit is used to iteratively evolve a target path planning scheme within the solution space using an initial path planning scheme as the initial population. The evolutionary process is configured such that in each generation of evolution, a first optimization objective is obtained based on various resource consumption data, a second optimization objective is obtained based on various weight priorities, and operations on path nodes are performed.

[0060] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0061] Accordingly, embodiments of the present invention provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform steps in the path planning method of the above embodiments, for example... Figure 1 Steps S1 to S6 as described above.

[0062] This invention, through its embodiments, creatively introduces a dual dynamic evaluation mechanism by constructing a basic framework for acquiring traffic path data and location coordinate data of each path node in a target area and generating an initial path planning scheme. On one hand, based on the real-time status data of each acquired path node, the initial weight priority of each path node is determined. On the other hand, the acquired user information is analyzed and processed, and the initial weight priorities are adjusted based on the processing results to obtain individual weight priorities, thereby transforming static path nodes into intelligent nodes with dynamic business value. Furthermore, the scheme further breaks down the traffic path data based on each path node and determines the resource consumption data for each traffic path. Finally, using multi-objective optimization technology, with user constraint information as constraints, the initial path planning scheme is corrected based on the resource consumption data and the weight priorities to obtain the target path planning scheme. The embodiments of the present invention no longer solely pursue the minimization of physical distance. Instead, under the premise of strictly adhering to user constraints, they intelligently weigh the weights and priorities reflecting the importance of nodes, automatically avoid highly congested or low-value road segments, and prioritize connecting high-priority task points. This results in the output of the optimal path that satisfies both the user's personalized execution preferences and the real-time load of the region, effectively improving the comfort of task execution and the scientific nature of resource scheduling.

[0063] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A path planning method, characterized in that, include: Acquire the travel path data of the target area and the location coordinate data of each path node in the target area; The travel path data and the coordinate data of each location are input into the pre-constructed path planning model to obtain the initial path planning scheme; Based on the real-time status data of each path node, the initial weight priority of each path node is determined. The travel path data is split based on each of the path nodes, and the resource consumption data of each travel path is determined. The acquired user information is analyzed and processed, and the initial weight priorities are adjusted based on the processing results to obtain the respective weight priorities; Using multi-objective optimization techniques and user constraint information as constraints, the initial path planning scheme is modified based on the resource consumption data and the weights and priorities of each resource to obtain the target path planning scheme.

2. The path planning method as described in claim 1, characterized in that, The determination of the initial weight priority of each path node based on the acquired real-time status data of each path node includes: The corresponding queuing time and the running data of the path node are extracted from each of the real-time status data. Based on each of the aforementioned queuing times, a first weighted priority sequence is obtained; Based on the aforementioned operational data, a second weight priority sequence is obtained; Based on the first weight priority sequence and the second weight priority sequence, the initial weight priority of each path node is obtained.

3. The path planning method as described in claim 1, characterized in that, The step of splitting the travel path data based on each of the path nodes and determining the resource consumption data for each travel path includes: The travel path data is split based on each of the path nodes to obtain the travel path between any two of the path nodes; For each of the aforementioned travel paths, the corresponding resource consumption data is obtained based on the determined travel method.

4. The path planning method as described in claim 1, characterized in that, The process involves analyzing and processing the acquired user information, and adjusting the initial weight priorities based on the processing results to obtain various weight priorities, including: The acquired user information is processed to determine the value of interest for each path node; Based on the user information, the data on consumable resources is determined; The initial weight priority is adjusted based on the consumable resource data and each of the values ​​of interest to obtain the respective weight priority.

5. The path planning method as described in claim 1, characterized in that, The method employs multi-objective optimization technology, using user constraint information as constraints, and modifies the initial path planning scheme based on the resource consumption data and the weight priorities to obtain the target path planning scheme, including: Parse the user constraint information and construct the solution space; Using the initial path planning scheme as the initial population, the target path planning scheme is obtained by iterative evolution using a multi-objective optimization algorithm within the solution space. The evolution process is configured such that in each generation of evolution, a first optimization objective is obtained based on each of the resource consumption data, a second optimization objective is obtained based on each of the weight priorities, and operations are performed on the path nodes.

6. A path planning system, characterized in that, include: The acquisition module is used to acquire the travel path data of the target area and the position coordinate data of each path node in the target area; The initial scheme module is used to input the travel path data and the coordinate data of each location into a pre-constructed path planning model to obtain an initial path planning scheme; The initial priority module is used to determine the initial weight priority of each path node based on the real-time status data of each path node obtained. The splitting module is used to split the travel path data based on each of the path nodes and determine the resource consumption data of each travel path. The analysis module is used to analyze and process the acquired user information, and adjust the initial weight priority based on the processing results to obtain the weight priority. The correction module is used to employ multi-objective optimization techniques, taking user constraint information as constraints, and correcting the initial path planning scheme based on the resource consumption data and the weight priorities of each resource to obtain the target path planning scheme.

7. The path planning system as described in claim 6, characterized in that, The initial priority module includes: An extraction unit is used to extract the corresponding queuing time and the running data of the path node from each of the real-time status data. The first unit is used to obtain a first weighted priority sequence based on each of the queuing times; The second unit is used to obtain a second weight priority sequence based on each of the aforementioned operational data; A priority unit is used to obtain the initial weight priority of each of the path nodes based on the first weight priority sequence and the second weight priority sequence.

8. The path planning system as described in claim 6, characterized in that, The splitting module includes: A travel path unit is used to split the travel path data based on each of the path nodes to obtain the travel path between any two of the path nodes; The resource consumption data unit is used to obtain the corresponding resource consumption data for each travel path based on a determined travel method.

9. The path planning system as described in claim 6, characterized in that, The analysis module includes: The interest value unit is used to process the acquired user information and determine the interest value of each path node; A consumable resource unit is used to determine consumable resource data based on the user information. An adjustment unit is used to adjust the initial weight priority based on the consumable resource data and each of the values ​​of interest, to obtain each weight priority.

10. The path planning system as described in claim 6, characterized in that, The correction module includes: The parsing unit is used to parse the user constraint information and construct the solution space; An evolutionary unit is used to iteratively evolve the initial path planning scheme as the initial population within the solution space using a multi-objective optimization algorithm to obtain the target path planning scheme. The evolutionary process is configured to obtain a first optimization objective based on each of the resource consumption data and a second optimization objective based on each of the weight priorities in each generation of evolution, and to perform operations on the path nodes.

Images