Reliable path planning method and system based on travel time prediction

Abstract

The invention discloses a reliable path planning method and system based on travel time prediction, and relates to the technical field of intelligent traffic and path planning, and the method comprises the steps: inputting a vehicle track, departure time, weather and other information into a diffusion model, obtaining the probability distribution of the travel time of a road section, constructing a random time-varying road network, and carrying out the prediction of the travel time; and defining a fastest path by using a travel time quantile, embedding a quantile index into an improved A * search to construct a path scoring function, and outputting a path meeting on-time arrival probability constraint. According to the method, the random time-varying characteristic of the travel time can be described without a preset distribution form, and the long-tail risk in peak hours and severe weather is reflected; the candidate path scale is compressed through path control pruning based on quantiles, and the distribution calculation overhead is reduced; in combination with an improved A * search framework, a reliable fastest path is rapidly solved in a large-scale road network, and the on-time arrival reliability and the path planning efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation and route planning technology, and in particular to a reliable route planning method and system based on travel time prediction. Background Technology

[0002] With rapid economic and social development, the ever-increasing number of motor vehicles and their mileage has led to severe traffic congestion, resulting in significantly longer travel times. Against this backdrop, finding routes that save travel time, avoid severe congestion, and ensure on-time arrival has become a necessity for daily travel. With the widespread adoption of mobile internet, mobile navigation apps on smartphones have become ubiquitous tools for travel, and route planning is one of the core services offered by these apps.

[0003] Current route planning methods mostly calculate the shortest route based on traffic conditions at the time of user query. However, since traffic conditions are constantly changing during travel, the traffic information at the user's departure time is inherently lagging. Ignoring the uncertainty of travel time leads to suboptimal calculated routes and fails to guarantee on-time arrival for travelers. More advanced methods describe the travel time of road segments probabilistically, but the random and time-varying travel times of road segments make the travel time distribution of the path difficult to calculate directly. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a reliable path planning method based on travel time prediction, which solves the problem of efficiently solving the reliable fastest path when considering time-varying factors and on-time arrival probability.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a reliable route planning method based on travel time prediction, which includes constructing a conditional diffusion model by using gridded vehicle travel trajectories and vehicle departure times, weather and holiday information, and obtaining a road segment travel time probability distribution based on the road segment travel time prediction sample set output by the diffusion model.

[0008] A stochastic time-varying road network model is constructed using the probability distribution of travel time for road segments. In this model, the probability distribution of travel time for each road segment is determined by randomly superimposing the travel times of each road segment.

[0009] The travel time quantiles are determined based on the probability distribution of travel time along the path; the reliable fastest path and the path dominance relationship are defined using the travel time quantiles, and dominated paths are eliminated.

[0010] A path scoring function is constructed based on the path travel time quantiles. An improved A* search algorithm is used to expand the path with the smallest scoring function value. Finally, the reliable fastest path from the starting node to the destination node is determined by utilizing the path dominance relationship.

[0011] As a preferred embodiment of the reliable path planning method based on travel time prediction described in this invention, the probability distribution of travel time for road segments includes dividing the road network area into spatial grids, constructing a spatial grid sequence based on vehicle travel trajectories, and inputting the spatial grid sequence, departure time, weather, and holiday information into a representation learning network to obtain a feature representation of the vehicle travel trajectory.

[0012] Based on the trajectory feature representation, a conditional diffusion model is constructed. A noise-adding and noise-reducing process is performed on the historical travel time. Under the condition of a given departure time, the conditional diffusion model is randomly sampled to obtain a travel time prediction sample set for each road segment. The sample set is then used as a statistical sample to calculate the probability distribution of the travel time for each road segment.

[0013] As a preferred embodiment of the reliable path planning method based on travel time prediction described in this invention, the stochastic time-varying road network model includes randomly superimposing the probability distributions of the travel times of the road segments to construct the stochastic time-varying road network model.

[0014] The day is divided into time slices according to a preset time granularity. In the t-th time slice... Inside, through the section of road The travel time is treated as a random variable, and the probability distribution of the travel time of a road segment is denoted as... .

[0015] For any path whose starting point is node r At the departure time When the time is, the travel time from node r to the next node j via intermediate node i is expressed as the sum of the travel time from node r to node i and the travel time from node i to the route segment ij.

[0016] The probability distribution of the corresponding path travel time:

[0017]

[0018] in, Indicates in Departing at time, following path u from node Departure to Node The probability distribution of travel time; Indicates in Departure time indicates that Departing at time, following path u from node The probability distribution of travel time from departure to node i; Indicates in The probability distribution of the travel time for a journey along road segment ij, starting from a given time.

[0019] As a preferred embodiment of the reliable path planning method based on travel time prediction described in this invention, the definition of the reliable fastest path includes, in a stochastic time-varying road network model, given the departure time... and on-time arrival probability ,remember It is the set of all feasible paths from the starting node r to the destination node s.

[0020] For any candidate path Based on the probability distribution of travel time Determine the travel time quantiles of the route. .

[0021] If a path exists This allows for the same departure time and on-time arrival probability Under the condition that the travel time quantile of a path is less than the travel time quantile of any other path, then the path will be... Defined as the reliable fastest path from node r to node s.

[0022]

[0023] in, Indicates the departure time is The probability of arriving on time is Under the given conditions, starting from the starting node r, along the reliable fastest path Travel time distribution to destination node s The corresponding travel time quantiles; Indicates the departure time is The probability of arriving on time is Under the given conditions, starting from the starting node r, along the candidate path Travel time distribution to destination node s The corresponding travel time quantiles.

[0024] As a preferred embodiment of the reliable route planning method based on travel time prediction described in this invention, the route dominance relationship includes defining the route dominance relationship based on travel time quantiles and eliminating dominated paths.

[0025] In a stochastic time-varying road network model, the starting node is determined. To the intermediate node Two partial paths ;in, Indicates from node To the node The set of all feasible paths, Indicates in set In the middle, the first Article from Feasible paths to i; Indicates in set In the middle, the first Article from arrive A feasible path.

[0026] path , Expand to slave node To any node full path If the probability of arriving on time is Under the condition of node Along the path Reaching the node The travel time quantile is smaller than that along the path. The path is determined by the quantiles of the travel time. Dominant path in stochastic time-varying road networks and the dominated part of the path Remove.

[0027]

[0028] in, Indicates the departure time is The probability of arriving on time is Under the condition of starting node along the path Travel time distribution to node w The corresponding travel time quantiles; Indicates the departure time is The probability of arriving on time is Under the condition of starting node along the path Reaching the node Travel time distribution The corresponding travel time quantiles.

[0029] As a preferred embodiment of the reliable path planning method based on travel time prediction described in this invention, the construction of the path scoring function includes, in the stochastic time-varying road network model, setting the path sequence to be processed (SE) to store the path from the starting node. Candidate paths to each node.

[0030] The path sequence SE to be processed is initialized using the labeling method, and a starting node is created. The initial path to itself .

[0031] For the path sequence SE to be processed, the departure time is The probability of arriving on time is Under the given conditions, for each partial path from the starting node r to node i We calculate the travel time quantiles of the path and introduce a heuristic distance function from node i to destination node s to construct a path scoring function.

[0032]

[0033] in, This represents the path from the starting node r to the current node i. The calculated path score function value; Indicates the departure time is The probability of arriving on time is Under the condition that, from node r along the path The quantiles of the travel time to node i; The heuristic function representing the distance from the current node i to the destination node s.

[0034] As a preferred embodiment of the reliable path planning method based on travel time prediction described in this invention, the improved A* search algorithm includes, in each iteration, selecting the path with the smallest score function value from the path sequence SE to be processed as the current extended path, based on the path score function. And set node i as the current extended node.

[0035] For each neighbor node j of the current expanding node i, along the road segment path Extending the path, we obtain a new path from the starting node r to node j. .

[0036]

[0037] in, This represents the u-th path from the starting node r to the current node i; This represents the u-th path from the starting node r to the current node i; This represents the road segment from node i to node j; This represents a path concatenation operator that connects the original path with the road segment.

[0038] Using the path dominance determination method, the new path The new path is compared with other candidate paths to node j in the path sequence SE to be processed; if the new path is dominated by any candidate path, the new path is discarded; if the new path is not dominated by any candidate path, the new path is added to the path sequence SE to be processed, and the candidate paths dominated by the new path are deleted.

[0039] Repeat the path selection, path expansion, and path dominance determination process until the endpoint node of the current expanded path selected from the path sequence SE is the destination node s; determine the current expanded path as the reliable fastest path from the starting node r to the destination node s, and use it as the path planning result.

[0040] Secondly, the present invention provides a reliable route planning system based on travel time prediction, including a gridding module: constructing a conditional diffusion model through gridded vehicle travel trajectories and vehicle departure times, weather and holiday information, and obtaining the probability distribution of road segment travel time based on the road segment travel time prediction sample set output by the diffusion model.

[0041] Random Time-Varying Road Network Module: A random time-varying road network model is constructed using the probability distribution of road segment travel time. In the model, the probability distribution of the travel time of the path is determined by randomly superimposing the travel times of each road segment.

[0042] Definition module: Determine the travel time quantiles based on the travel time probability distribution of the path; use the travel time quantiles to define the reliable fastest path and the path dominance relationship, and eliminate dominated paths.

[0043] Scoring Module: Constructs a path scoring function based on the path travel time quantiles, expands the path with the smallest scoring function value using an improved A* search algorithm, and determines the reliable fastest path from the starting node to the destination node using path dominance relationships.

[0044] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the reliable path planning method based on travel time prediction as described in the first aspect of the present invention.

[0045] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the reliable path planning method based on travel time prediction as described in the first aspect of the present invention.

[0046] The beneficial effects of this invention are as follows: By constructing a conditional diffusion model based on vehicle travel trajectory, departure time, weather, and holiday information, a fine-grained travel time probability distribution can be obtained at the road segment scale, thereby forming a stochastic time-varying road network model that more closely reflects the actual operating conditions; by introducing a travel time quantile index at the path level, reliable fastest paths and path dominance relationships are defined, eliminating a large number of dominated candidate paths while ensuring a preset on-time arrival probability, effectively reducing the search space; based on this, the quantile index is embedded into an improved A* search algorithm to construct a path scoring function, enabling the path planning process to simultaneously consider traffic efficiency and on-time arrival reliability, significantly improving the computational efficiency of finding reliable fastest paths in large-scale road networks and the robustness of path solutions. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of a reliable route planning method based on travel time prediction.

[0049] Figure 2 This is a schematic diagram of a road network for a reliable path planning method based on travel time prediction.

[0050] Figure 3 This is a graph showing the relationship between the probability of on-time arrival and the estimated travel time for a reliable route planning method based on travel time prediction. Detailed Implementation

[0051] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0052] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0053] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0054] Reference Figures 1-3 This is one embodiment of the present invention, which provides a reliable route planning method based on travel time prediction, including the following steps:

[0055] S1: A conditional diffusion model is constructed using gridded vehicle travel trajectories, vehicle departure times, weather, and holiday information. Based on the road segment travel time prediction sample set output by the diffusion model, the probability distribution of road segment travel time is obtained.

[0056] By gridding the road network area and constructing a spatial grid sequence based on vehicle travel trajectories, spatial location information, along with factors such as departure time, weather, and date, is input into a representation learning network that integrates self-attention and cross-attention. This allows for a precise characterization of the joint dependence of road segment travel time on spatial proximity and time-varying environmental factors. Furthermore, a diffusion model conditioned on trajectory feature representation is introduced to perform noise addition and denoising processes on travel time. By performing multiple random samplings of Gaussian random variables to obtain a travel time prediction sample set, the probability distribution of road segment travel time can be directly modeled based on data without pre-assuming a specific distribution form. This better reflects the long-tail and multi-peak characteristics of travel time under scenarios such as morning and evening rush hours and severe weather, providing more accurate and robust travel time probability inputs for subsequent construction of stochastic time-varying road network models and reliable path planning based on quantiles.

[0057] The probability distribution of travel time for the road segment includes dividing the road network area into spatial grids, constructing a spatial grid sequence based on the vehicle's travel trajectory, and inputting the spatial grid sequence, departure time, weather, and holiday information into a representation learning network to obtain a feature representation of the vehicle's travel trajectory.

[0058] Based on the trajectory feature representation, a conditional diffusion model is constructed. A noise-adding and noise-reducing process is performed on the historical travel time. Under the condition of a given departure time, the conditional diffusion model is randomly sampled to obtain a travel time prediction sample set for each road segment. The sample set is then used as a statistical sample to calculate the probability distribution of the travel time for each road segment.

[0059] In the noise addition stage, the travel time obtained based on historical vehicle trajectories is used as the base value. In each iteration, Gaussian random noise with a mean of 0 and a variance of 1 is added to the base value to form a Gaussian random variable through multiple iterations. In the noise removal stage, the characteristic representation of the vehicle trajectories is used as the conditional input to estimate the noise components added in each iteration. The Gaussian random variable is then gradually denoised in the reverse order of the noise addition process to obtain the predicted travel time value.

[0060] Specifically, 50 steps of noise addition are performed in the noise addition stage of the diffusion model, and 100 Gaussian noises are randomly sampled in the denoising stage, so that the probability distribution is composed of 100 travel time prediction results. Compared with other mainstream methods, the experimental results show that the prediction accuracy of the proposed method is improved by 5%-10%.

[0061] S2: Construct a stochastic time-varying road network model using the probability distribution of road segment travel time. In the model, the probability distribution of the travel time of the path is determined by randomly superimposing the travel times of each road segment.

[0062] By dividing a day into several time slices, treating the travel time of road segments as random variables within each time slice, and progressively superimposing the travel time distribution from the starting node to the intermediate node with the travel time distribution of continuing along the next road segment at the intermediate node during path expansion, the random and time-varying characteristics of each road segment can be naturally inherited at the path level. This recursive modeling approach from road segment to path avoids the information loss caused by simply using average time to replace probability distribution, and more accurately portrays uncertainties such as peak periods and congestion. On the other hand, it facilitates real-time updating of the travel time probability distribution of candidate paths during the search process, providing a unified and easily computed model foundation for subsequent path dominance determination based on quantiles and reliable fastest path solutions.

[0063] like Figure 2 The diagram shown is a schematic of the random time-varying road network used in this invention. Nodes 1 to 24 represent various locations or intersections in the road network. The arrowed lines between nodes are directed road segments, and the numbers 1 to 76 on the side are road segment numbers used to index their travel time probability distribution.

[0064] The stochastic time-varying road network model includes randomly superimposing the probability distributions of the travel time of the road segments to construct the stochastic time-varying road network model.

[0065] The day is divided into time slices according to a preset time granularity. In the t-th time slice... Inside, through the section of road The travel time is treated as a random variable, and the probability distribution of the travel time of a road segment is denoted as... .

[0066] For originating point as node any path At the departure time At that time, from the node Passing through intermediate nodes Reach the next node The travel time is expressed as the time from node To the node Travel time and at the node Starting point and route The sum of the travel times.

[0067] The probability distribution of the corresponding path travel time:

[0068]

[0069] in, Indicates in Departing at time, following path u from node Departure to Node The probability distribution of travel time; Indicates in Departure time indicates that Departing at time, following path u from node The probability distribution of travel time from departure to node i; Indicates in The probability distribution of the travel time for a journey along road segment ij, starting from a given time.

[0070] S3: Determine the travel time quantiles based on the travel time probability distribution of the path; use the travel time quantiles to define the reliable fastest path and the path dominance relationship, and eliminate dominated paths.

[0071] In stochastic time-varying road network models, given the departure time and on-time arrival probability, the minimum travel time quantile is used as the criterion for determining the reliable fastest path. This makes the fastest path no longer simply the shortest average travel time, but rather the path that requires the least time to reach the preset on-time arrival probability. Thus, traffic efficiency and on-time reliability are unified under the same evaluation index. Compared to using only the expected time or the shortest path length as the optimization objective, this definition can more sensitively reflect the tail risk in scenarios such as peak periods, accidents, and severe weather, making the selected path more robust under the preset on-time arrival probability constraint.

[0072] The definition of the reliable fastest path includes, in a stochastic time-varying road network model, given a departure time and on-time arrival probability ,remember Let r be the set of all feasible paths from the starting node r to the destination node s; for any candidate path Based on the probability distribution of travel time Determine the travel time quantiles of the route. .

[0073] If a path exists This allows for the same departure time and on-time arrival probability Under the condition that the travel time quantile of a path is less than the travel time quantile of any other path, then the path will be... Defined as the reliable fastest path from node r to node s.

[0074]

[0075] in, Indicates the departure time is The probability of arriving on time is Under the given conditions, starting from the starting node r, along the reliable fastest path Travel time distribution to destination node s The corresponding travel time quantiles; Indicates the departure time is The probability of arriving on time is Under the given conditions, starting from the starting node r, along the candidate path Travel time distribution to destination node s The corresponding travel time quantiles.

[0076] By establishing path dominance relationships based on travel time quantiles: For partial paths from the starting node to intermediate nodes, pairwise comparisons are performed. If a partial path exists whose travel time quantile is not inferior to another partial path regardless of its extension to any subsequent node, the latter is considered the dominated path and eliminated. This design allows for the elimination of a large number of candidate paths that are simultaneously inferior in reliability and efficiency in the early stages of path expansion, effectively reducing the subsequent search space. While ensuring the retention of reliable fastest paths, it reduces invalid path expansion and distributed computation, thereby significantly improving the computational efficiency of finding reliable fastest paths in large-scale road networks.

[0077] The path dominance relationship includes defining the path dominance relationship based on the travel time quantile and eliminating the dominated paths.

[0078] In a stochastic time-varying road network model, the starting node is determined. To the intermediate node Two partial paths ;in, Indicates from node To the node The set of all feasible paths, Indicates in set In the middle, the first Article from Feasible paths to i; Indicates in set In the middle, the first Article from arrive A feasible path.

[0079] path , Expand to slave node To any node full path If the probability of arriving on time is Under the condition of node Along the path Reaching the node The travel time quantile is smaller than that along the path. The path is determined by the quantiles of the travel time. Dominant path in stochastic time-varying road networks and the dominated part of the path Remove.

[0080]

[0081] in, Indicates the departure time is The probability of arriving on time is Under the condition of starting node along the path Travel time distribution to node w The corresponding travel time quantiles; Indicates the departure time is The probability of arriving on time is Under the condition of starting node along the path Reaching the node Travel time distribution The corresponding travel time quantiles.

[0082] Based on path dominance, the on-time arrival probability interval (0,1) is discretized as follows: Each value point, for the path The passage time quantile function is discretized.

[0083]

[0084] in, This represents the discrete on-time arrival probability value used to determine path dominance; This represents the distance from the probability interval (0,1) for on-time arrival. , This indicates the number of parts into which the interval is divided.

[0085] Calculate the path at each discrete point where the probability of on-time arrival is taken. , The transit time quantile; if for all discrete values All satisfy:

[0086]

[0087] in, Indicates the departure time is The probability of arriving on time is Under the condition of starting node Along the path Reaching the node Travel time distribution The corresponding transit time quantile; This indicates the departure time is The probability of arriving on time is Under the condition of starting node Along the path Reaching the node Travel time distribution The corresponding quantile of travel time.

[0088] If the above conditions are met, then it is determined that the departure time is Under the condition of starting node Path to intermediate node i Dominant path in stochastic time-varying road networks and the dominated part of the path From the candidate path set Remove from the list.

[0089] S4: Construct a path scoring function based on the path travel time quantiles, use an improved A* search algorithm to expand the path with the smallest scoring function value, and use path dominance to determine the reliable fastest path from the starting node to the destination node.

[0090] By setting the path sequence SE to be processed in the stochastic time-varying road network model and initializing the initial path from the starting node to itself using the labeling method, a path scoring function based on the travel time quantile and heuristic distance is introduced for each partial path. This allows the current travel time level under the preset on-time arrival probability constraint and the estimated remaining cost from the current node to the destination node to be reflected in the same evaluation index. This balances the reliability of the path and the global search directionality, avoiding blind expansion based solely on local shortest time or geometric distance.

[0091] During path expansion, an improved A* search algorithm is employed to select the path with the smallest scoring function from the SE for expansion. After generating new paths for all neighboring nodes of the current expansion node, the path dominance relationship based on travel time quantiles is combined with the aforementioned dominance judgment and pruning of multiple candidate paths reaching the same node. This ensures that more reliable and efficient candidate paths are always prioritized for expansion, while also eliminating a large number of paths that are inferior in terms of quantiles in the early stages of the search, significantly reducing the size of the sequence to be processed. Through this improved A* search framework guided by scoring function, path dominance pruning, and destination-first popping, the efficient solution of reliable and fastest paths in large-scale stochastic time-varying road networks is achieved while ensuring that the final path satisfies the definition of reliable and fastest, significantly improving the computational efficiency and result quality of path planning.

[0092] like Figure 3 As shown, the estimated travel time of the reliable fastest path increases with the increase of the probability of on-time arrival; experimental results show that the improved A* algorithm takes 23ms and the labeling algorithm takes 29ms. By comparison, the efficiency of the improved A* algorithm based on the dominant path is improved by 20%.

[0093] The constructed path scoring function includes, in the stochastic time-varying road network model, setting the path sequence to be processed (SE) to store the path from the starting node. Candidate paths to each node.

[0094] The path sequence SE to be processed is initialized using the labeling method, and a starting node is created. The initial path to itself .

[0095] For the path sequence SE to be processed, the departure time is The probability of arriving on time is Under the condition, for each starting node Partial path to node i Calculate the travel time quantiles of the path and introduce node i to the destination node. A heuristic distance function is used to construct a path scoring function.

[0096]

[0097] in, This represents the path from the starting node r to the current node i. The calculated path score function value; Indicates the departure time is The probability of arriving on time is Under the condition that, from node r along the path The quantiles of the travel time to node i; The heuristic function representing the distance from the current node i to the destination node s.

[0098] The improved A* search algorithm includes, based on a path scoring function, selecting the path with the smallest scoring function value from the path sequence SE to be processed as the current expanded path in each iteration. And set node i as the current extended node.

[0099] For each neighbor node of the current expanding node i Along the road section path Extending the path, we obtain a new path from the starting node r to node j. .

[0100]

[0101] in, This represents the u-th path from the starting node r to the current node i; This represents the u-th path from the starting node r to the current node i; This represents the road segment from node i to node j; This represents a path concatenation operator that connects the original path with the road segment.

[0102] Using the path dominance determination method, the new path The new path is compared with other candidate paths to node j in the path sequence SE to be processed; if the new path is dominated by any candidate path, the new path is discarded; if the new path is not dominated by any candidate path, the new path is added to the path sequence SE to be processed, and the candidate paths dominated by the new path are deleted.

[0103] Repeat the path selection, path expansion, and path dominance determination process until the endpoint node of the currently expanded path selected from the path sequence SE is the destination node. The current extended path is determined to start from the originating node. to the destination node The reliable fastest path is used as the result of path planning.

[0104] This embodiment also provides a reliable route planning system based on travel time prediction, including: a gridding module: constructing a conditional diffusion model through gridded vehicle travel trajectories and departure times, weather and holiday information, and obtaining the probability distribution of road segment travel times based on the travel time prediction sample set output by the diffusion model.

[0105] Random Time-Varying Road Network Module: Constructs a random time-varying road network model using the probability distribution of road segment travel time. By randomly superimposing the travel times of each road segment on the path, the probability distribution of the travel time of the path is determined.

[0106] Definition module: Determine the travel time quantiles based on the travel time probability distribution of the path; use the travel time quantiles to define the reliable fastest path and the path dominance relationship, and eliminate dominated paths.

[0107] Scoring Module: Constructs a path scoring function based on the path travel time quantile, uses an improved A* search algorithm for path expansion, and utilizes path dominance relationships to determine the reliable fastest path from the starting node to the destination node.

[0108] This embodiment also provides a computer device applicable to the reliable path planning method based on travel time prediction, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the reliable path planning method based on travel time prediction as proposed in the above embodiment.

[0109] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0110] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the reliable path planning method based on travel time prediction as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0111] In summary, this invention constructs a conditional diffusion model using vehicle travel trajectories and information such as departure time, weather, and holidays to obtain the probability distribution of road segment travel time and establish a stochastic time-varying road network model. Based on this, it defines reliable fastest paths and path dominance relationships based on travel time quantiles and embeds quantile indices into an improved A* search algorithm to construct a path scoring function. This achieves efficient solution for reliable fastest paths in large-scale road networks while satisfying preset on-time arrival probability constraints, thereby more accurately characterizing the stochastic time-varying characteristics of travel time, effectively compressing the size of candidate paths, and improving the on-time reliability and computational efficiency of path planning results.

[0112] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A reliable path planning method based on travel time prediction, characterized by: The road section travel time probability distribution comprises: dividing a road network region into spatial grids, constructing a spatial grid sequence according to vehicle travel trajectories, and inputting the spatial grid sequence, a departure time, weather and holiday information into a representation learning network to obtain a vehicle travel trajectory feature representation; The random time-varying road network model comprises: performing random superposition on the probability distribution of the road section travel time to construct the random time-varying road network model; The path domination relationship comprises: defining a path domination relationship based on the travel time quantile, and eliminating dominated paths; The path scoring function is constructed based on the path travel time quantile, the improved A* search algorithm is used to expand the path with the minimum scoring function value, and the reliable fastest path from the departure node to the destination node is determined by using the path domination relationship. 2.The reliable path planning method based on travel time prediction of claim 1, wherein: The road section travel time probability distribution comprises: dividing a road network region into spatial grids, constructing a spatial grid sequence according to vehicle travel trajectories, and inputting the spatial grid sequence, a departure time, weather and holiday information into a representation learning network to obtain a vehicle travel trajectory feature representation; The random time-varying road network model comprises: performing random superposition on the probability distribution of the road section travel time to construct the random time-varying road network model; 3.The reliable path planning method based on travel time prediction of claim 2, wherein: The path domination relationship comprises: defining a path domination relationship based on the travel time quantile, and eliminating dominated paths; A day is divided into time slices according to a preset time granularity, and the travel time of a road segment is regarded as a random variable in the tth time slice The probability distribution of the road segment travel time is denoted as ;​ For any path with origin node r When the departure time The travel time from node r to the next node j via an intermediate node i is represented as the sum of the travel time from node r to node i and the travel time over link ij from node i. The path scoring function is constructed based on the path travel time quantile, the improved A* search algorithm is used to expand the path with the minimum scoring function value, and the reliable fastest path from the departure node to the destination node is determined by using the path domination relationship. in, Indicates in Departing at time, following path u from node Departure to Node The probability distribution of travel time; Indicates in Departure time indicates that Departing at time, following path u from node The probability distribution of travel time from departure to node i; Indicates in The probability distribution of the travel time for a journey along road segment ij, starting from a given time.

4. The reliable path planning method based on travel time prediction of claim 3, wherein: The definition of the reliable fastest path includes, in a random time-varying road network model, given the departure time and the on-time arrival probability , record as the set of all feasible paths from the departure node r to the destination node s; For any candidate path , determine quantiles of travel times for the path based on the probability distribution of travel times ;​ If a path exists This allows for the same departure time and on-time arrival probability Under the condition that the travel time quantile of a path is less than the travel time quantile of any other path, then the path will be... Defined as the reliable fastest path from node r to node s; in, Indicates the departure time is The probability of arriving on time is Under the given conditions, starting from the starting node r, along the reliable fastest path Travel time distribution to destination node s The corresponding travel time quantiles; Indicates the departure time is The probability of arriving on time is Under the given conditions, starting from the starting node r, along the candidate path Travel time distribution to destination node s The corresponding travel time quantiles.

5. The reliable path planning method based on travel time prediction of claim 4, wherein: The road section travel time probability distribution comprises: dividing a road network region into spatial grids, constructing a spatial grid sequence according to vehicle travel trajectories, and inputting the spatial grid sequence, a departure time, weather and holiday information into a representation learning network to obtain a vehicle travel trajectory feature representation; In a stochastic time-varying road network model, two partial paths from a start node to an intermediate node are determined; wherein, denotes a set of all feasible paths from a node to a node , denotes the i-th feasible path from to i in the set , denotes the i-th feasible path from to in the set , denotes the i-th feasible path from to . The path , is extended to the complete path from the node to any node ; if the travel time quantile of the node from the node along the path is less than the travel time quantile of the node along the path , the path is determined to dominate the path in the stochastic time-varying road network, and the dominated part of the path is pruned. wherein, denotes the travel time quantile corresponding to the travel time distribution of the travel time from the departure node to the destination node w along the path with a departure time and a punctual arrival probability ; denotes the travel time quantile corresponding to the travel time distribution of the travel time from the departure node to the destination node along the path with a departure time and a punctual arrival probability .

6. The reliable path planning method based on travel time prediction of claim 5, wherein: The constructed path scoring function includes, in the stochastic time-varying road network model, setting the path sequence to be processed (SE) to store the path from the starting node. Candidate paths to each node; The label setting method is used to initialize the path sequence SE to be processed, and an initial path from the starting node to itself is created ; For the path sequence SE to be processed, at the departure time and the on-time arrival probability Under the condition, for each partial path from the departure node r to the node i , the travel time quantile of the path is calculated, and the heuristic distance function of the node i to the destination node s is introduced to construct the path scoring function; in, This represents the path from the starting node r to the current node i. The calculated path score function value; Indicates the departure time is The probability of arriving on time is Under the condition that, from node r along the path The quantiles of the travel time to node i; The heuristic function representing the distance from the current node i to the destination node s.

7. The reliable path planning method based on travel time prediction of claim 6, wherein: The improved A* search algorithm includes, based on a path scoring function, selecting the path with the smallest scoring function value from the path sequence SE to be processed as the current expanded path in each iteration. And set node i as the current extended node; For each neighbor node j of the current extended node i, the path is extended to obtain a new path from the start node r to node j ; and the node j is added to the path. wherein, denotes the u-th path from the start node r to the current node i; denotes the u-th path from the start node r to the current node i; denotes the link from node i to node j; denotes a path concatenation operator that connects the original path with the link. Using the path dominance determination method, the new path The new path is compared with other candidate paths to node j in the path sequence SE to be processed; if the new path is dominated by any candidate path, the new path is discarded; if the new path is not dominated by any candidate path, the new path is added to the path sequence SE to be processed, and the candidate paths dominated by the new path are deleted. The random time-varying road network model comprises: performing random superposition on the probability distribution of the road section travel time to construct the random time-varying road network model; 8. A reliable path planning system based on travel time prediction, based on the reliable path planning method based on travel time prediction according to any one of claims 1 to 7, characterized in that: The path domination relationship comprises: defining a path domination relationship based on the travel time quantile, and eliminating dominated paths; The path scoring function is constructed based on the path travel time quantile, the improved A* search algorithm is used to expand the path with the minimum scoring function value, and the reliable fastest path from the departure node to the destination node is determined by using the path domination relationship. ​ ​ 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The computer program is executed by the processor to implement the steps of the reliable path planning method based on travel time prediction according to any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the reliable path planning method based on travel time prediction according to any one of claims 1-7.

Images