Traffic dynamic shortest path distribution method and system based on triangular fuzzy number weight

Abstract

The invention discloses a traffic dynamic shortest path distribution method and system based on a triangular fuzzy number weight, and relates to the technical field of intelligent traffic systems.Fuzzy information in a path evaluation attribute is represented through a triangular fuzzy number, a comprehensive weight is analyzed and calculated in combination with deviation maximization and a moisture value, and the decision reliability is effectively improved; a K-Means algorithm is adopted to perform dynamic path search, and through a representative vertex subset pre-calculation and increment updating strategy, the path retrieval efficiency is remarkably improved; a travel time prediction model related to departure time and real-time flow is established, and accurate response to accidental congestion is realized; and finally, outputting an optimal path scheme through a multi-objective decision-making method. Compared with a traditional algorithm, the speed is improved on a real road network data set, the path distribution accuracy is improved, meanwhile, the path safety, comfort and efficiency are effectively balanced, real-time path planning under a large-scale road network is supported, and an efficient solution is provided for intelligent traffic management.

Description

A Traffic Dynamic Shortest Path Allocation Method and System Based on Triangular Fuzzy Number Weights Technical Field

[0001] This invention relates to the field of intelligent transportation systems, and in particular to a method and system for dynamic shortest path allocation based on triangular fuzzy number weights. Background Technology

[0002] The shortest path problem in dynamic traffic applications is crucial in many industries. For example, it's a vital sub-problem in trip planning, logistics planning, traffic simulation, navigation, biomedicine, social networks, and location-based services. These systems require rapid, real-time computation of shortest paths from various points. While current shortest path computation methods and algorithms have significantly improved the performance of algorithms on static graphs, existing algorithms for dynamic traffic applications still face several significant challenges.

[0003] With the increasing complexity and dynamism of urban transportation networks, traditional static shortest path algorithms are struggling to meet the high demands for real-time performance, accuracy, and robustness in practical applications. Especially when dealing with dynamic shortest path allocation problems, existing technologies face the following significant challenges:

[0004] 1. Insufficient characterization of the dynamic and uncertain traffic environment: The travel time of road segments in real-world traffic networks is not fixed but is affected by a variety of factors such as traffic flow, weather conditions, and unforeseen events, exhibiting strong time-varying and random characteristics. Traditional deterministic shortest path algorithms cannot effectively characterize and cope with this uncertainty, resulting in the planned paths often not being optimal in real-world traffic environments, and may even fail due to deteriorating traffic conditions.

[0005] 2. Limitations of Fuzzy Information Processing in Multi-Attribute Decision Making: When choosing a route, drivers or decision-making systems typically need to comprehensively consider multiple attributes such as travel time, distance, cost, safety, and comfort. Many of these attributes, such as "road congestion" and "driving comfort," are qualitative indicators that are difficult to describe with precise numerical values ​​and possess inherent fuzziness. While fuzzy multi-attribute decision-making methods, especially those based on triangular fuzzy numbers, provide a framework for processing such fuzzy information and can relatively accurately characterize the fuzziness in decision-making, traditional fuzzy decision-making methods usually assume that decision-makers can be completely certain of their judgments, which is often difficult to achieve in practice. The decision-maker's judgment of the importance of attributes and the evaluation of the performance of a solution under different attributes are themselves fuzzy and uncertain. Existing methods, when determining attribute weights, often only focus on measuring the differences in importance between attributes, such as based on deviation maximization or information uncertainty, such as based on entropy methods. They lack a reliable weight determination mechanism that comprehensively considers the distinguishability between attributes and the uncertainty of solution evaluation. This can easily lead to decision results with low distinguishability, insufficient persuasiveness, or even inconsistency with actual needs.

[0006] 3. Computational Efficiency Bottleneck of the K-Shortest Path Problem in Dynamic Large-Scale Road Networks: To address the risk of a single shortest path failing due to unforeseen circumstances, finding the top K shortest paths (K-Shortest Paths, KSP) provides more alternative travel options and helps balance traffic flow in the road network. However, solving the KSP problem in large-scale traffic networks is typically a computationally complex task, such as an NP-hard problem. The problem becomes even more complex when considering the dynamic characteristics of the road network. Many existing KSP algorithms, such as the classic Yen algorithm, are primarily designed for static networks. When weights change, they often require extensive repetitive computation or global updates, making it difficult to meet the real-world needs of rapid path replanning in dynamic traffic environments. Although some research has attempted to parallelize the algorithm or propose approximate algorithms, their efficiency and adaptability still need improvement when dealing with dynamic scenarios where edge weights change frequently and locally. Even if only a small portion of the edge weights change, existing acceleration techniques may struggle to handle this efficiently, necessitating dynamic algorithms capable of incremental updates rather than full recalculation.

[0007] 4. The accuracy of path travel time prediction models in describing dynamic traffic flow needs improvement: The key to achieving dynamic shortest path allocation lies in accurately predicting the travel time of vehicles traversing each path at a specific departure time. Traditional traffic impedance models often simply represent travel time as a function of traffic flow, failing to fully reflect the dynamic loading process of traffic flow and the propagation impact of occasional congestion on road segment traffic flow. The lack of detailed consideration of abrupt changes in road traffic conditions, such as congestion formation and dissipation, and the failure to transform travel time into a dynamic function closely related to departure time, real-time traffic flow, and capacity, limits the accuracy of path travel time prediction, thus affecting the effectiveness of dynamic path allocation.

[0008] Therefore, existing methods for dynamic shortest path allocation in traffic have significant shortcomings in handling environmental uncertainties and fuzzy information, efficiently solving dynamic K-shortest paths, accurately predicting dynamic travel times, and scientifically determining multi-attribute weights. Thus, there is an urgent need for a novel path allocation method that can comprehensively process triangular fuzzy information, combine an efficient dynamic KSP search algorithm, and accurately reflect the dynamic characteristics of traffic flow, in order to improve the intelligence level and practical application effectiveness of dynamic traffic navigation and management. Summary of the Invention

[0009] In view of this, in order to solve the problems existing in the current dynamic shortest path allocation method, the purpose of this invention is to provide a dynamic shortest path allocation method and system based on triangular fuzzy number weights. This method processes fuzzy information through triangular fuzzy numbers and combines deviation and moisture content to comprehensively determine attribute weights, making path selection decisions more realistic and with higher discriminative power. Based on a K-Means preprocessing and dynamic update strategy, it enables rapid retrieval of the K shortest path in large-scale dynamic road networks. The algorithm's running time is significantly lower than traditional nDijsk and nBFS algorithms, meeting real-time requirements. Furthermore, it can dynamically adjust and optimize target weights according to real-time conditions, and the dynamic traffic flow loading and travel time prediction model more accurately reflects actual traffic state changes, improving the accuracy of path travel time prediction.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] In a first aspect, the present invention provides a method for dynamic shortest path allocation in traffic based on triangular fuzzy number weights, comprising the following steps:

[0012] The evaluation attribute values ​​of candidate paths in the traffic network are obtained, and each attribute value is represented as a triangular fuzzy number. A multi-attribute decision matrix based on triangular fuzzy numbers is constructed.

[0013] Based on the multi-attribute decision matrix, the discrimination weight of each attribute value in the evaluation result is calculated by using deviation maximization. The uncertainty weight of the attribute information is calculated by combining the entropy value of all path evaluation attribute values ​​under each attribute value. The attribute weight based on discrimination and the uncertainty weight of attribute information are then weighted and fused to obtain the comprehensive attribute weight.

[0014] Based on the comprehensive attribute weights, a representative subset of vertices is selected from the road network vertex set to pre-calculate the shortest path distance between all vertex pairs, and the shortest path set is quickly retrieved.

[0015] By discretizing time using the point of sudden change in road traffic conditions as nodes, and dynamically loading traffic flow based on the traffic impedance model, a travel time prediction function related to vehicle departure time, real-time road segment flow, and road segment capacity is established to obtain the predicted travel time of each path at different departure times.

[0016] The path is comprehensively evaluated based on the combined attribute weights, the shortest path set, and the predicted travel time. The optimal recommended path is assigned to the vehicle, and the path scheme and travel time information are output.

[0017] As a further aspect of the present invention, the triangular fuzzy number is represented as: ,in, The left endpoint of the fuzzy number, The center point of the fuzzy number, Let be the right endpoint of the fuzzy number; the membership function of the triangular fuzzy number is defined as a piecewise linear function, used to characterize the fuzziness of attribute values, and the membership function is:

[0018] ;

[0019] In the formula, Let be any candidate value within the universe of discourse; when or At that time, the membership degree is 0.

[0020] As a further aspect of the present invention, the calculation of the discrimination weight specifically includes: measuring the degree of difference of each attribute in the evaluation results through the deviation maximization model, and the greater the degree of difference, the higher the weight; the calculation of the uncertainty weight specifically includes: measuring the uncertainty of the evaluation information under each attribute through the entropy value, and the higher the uncertainty, the lower the weight adjustment factor; the weighted fusion adopts a linear weighting method, and the fusion coefficient is determined based on the reliability index.

[0021] As a further aspect of the present invention, when pre-calculating the shortest path distance between all vertex pairs, K-Means is used for path search, including the following steps:

[0022] Preprocessing stage: Select a subset R from the vertex set V as representative vertices, and pre-calculate the shortest path distance between all pairs of vertices in R;

[0023] Dynamic retrieval phase: When the weights of the road network edges change, the set of affected vertices is identified, and an incremental update strategy is used to locally correct the pre-calculated data;

[0024] Path generation stage: Combining the A* algorithm and candidate path management mechanism, K shortest paths are quickly retrieved using pre-calculated distances.

[0025] As a further aspect of the present invention, the incremental update strategy is as follows: when a change in edge weight is detected, only the shortest path subtree affected by the edge is recalculated, and the relevant candidate path set is updated to avoid recalculating the entire path network; the strategy maintains the set of affected vertices. To achieve, among which, Representing an edge The subset consisting of the endpoint and all its descendant nodes.

[0026] As a further aspect of the present invention, the travel time prediction function is: ,in, The vehicle's departure time. For real-time traffic flow, For the traffic capacity of the road segment; function It is obtained by fitting historical data or training by machine learning, and is used to reflect the dynamic changes in road resistance under occasional congestion; dynamic loading is based on a time discretization strategy, which divides continuous time into discrete intervals for processing.

[0027] As a further aspect of the present invention, when comprehensively evaluating the path based on the comprehensive attribute weights, the shortest path set, and the predicted travel time, and allocating the optimal recommended path to the vehicle, a multi-objective decision-making method is used to allocate the optimal path to the vehicle. The multi-objective decision-making includes at least one of the following optimization objectives: minimizing the total travel time, minimizing the path distance, equalizing traffic flow, and optimizing lane-changing safety and comfort. The multi-objective decision-making uses a triangular fuzzy controller to dynamically adjust the weights of each objective. The inputs of the controller include the lane-changing safety coefficient, the vehicle speed, and the road surface adhesion coefficient, and the output is the comfort weight.

[0028] As a further aspect of the present invention, the triangular fuzzy controller includes:

[0029] The universe of discourse for lane-changing safety factor, vehicle speed, and adhesion coefficient is divided into fuzzy subsets, and a fuzzy control rule table is formulated.

[0030] The comfort weights are output by defuzzification and used to adjust the weights of comfort indicators in path evaluation.

[0031] Secondly, the present invention also provides a traffic dynamic shortest path allocation system based on triangular fuzzy number weights, comprising:

[0032] The data acquisition module is used to obtain the topology of the traffic network, multiple evaluation attribute values ​​of each candidate path, real-time traffic flow data, and vehicle status information.

[0033] The weight calculation module is communicatively connected to the data acquisition module. It is used to represent each evaluation attribute value as a triangular fuzzy number to construct a multi-attribute decision matrix. Based on the decision matrix, it calculates the discrimination weight of each attribute using the deviation maximization method. It also calculates the uncertainty weight of each attribute using the entropy value and merges the two types of weights to obtain the comprehensive attribute weight.

[0034] The path search module is communicatively connected to the weight calculation module and is used to perform path search using the K-Means algorithm based on the comprehensive attribute weights. The path search includes: selecting a representative subset of vertices from the network vertex set and pre-calculating the shortest path distance between all vertex pairs; and using an incremental update strategy to locally correct the pre-calculated data when the weights of the network edges change, so as to quickly retrieve the top K shortest paths.

[0035] The travel time prediction module is communicatively connected to the data acquisition module. It is used to discretize time with the point of sudden change in road traffic state as the node, dynamically load traffic flow based on traffic impedance model, and model the travel time of road segment as a function related to vehicle departure time, real-time road segment flow and road segment capacity, so as to predict the dynamic travel time of each path.

[0036] The decision and output module is communicatively connected to the weight calculation module, the path search module, and the travel time prediction module, respectively. It is used to combine the comprehensive attribute weights, the set of K shortest paths, and the predicted dynamic travel time to allocate the optimal path to the vehicle through a multi-objective decision-making method, and output the path scheme and the estimated travel time information.

[0037] As a further aspect of the present invention, the triangular fuzzy number used to represent the evaluation attribute value in the weight calculation module is: ,in, The left endpoint of the fuzzy number, The center point of the fuzzy number, Let be the right endpoint of the fuzzy number; the membership function of the triangular fuzzy number is defined as a piecewise linear function, used to characterize the fuzziness of attribute values, and the membership function is:

[0038] ;

[0039] In the formula, Let be any candidate value within the universe of discourse; when or At that time, the membership degree is 0.

[0040] As a further aspect of the present invention, the path search module includes:

[0041] The preprocessing unit is used to select a subset R from the vertex set V as representative vertices and pre-calculate the shortest path distance between all pairs of vertices in R.

[0042] The dynamic processing unit is used to identify the set of affected vertices when the weights of the road network edges change, and to locally correct the pre-calculated data in the preprocessing unit using an incremental update strategy.

[0043] The path generation unit is used to combine the A* algorithm and the candidate path management mechanism to quickly retrieve K shortest paths using the pre-calculated distance.

[0044] Compared with existing technologies, the traffic dynamic shortest path allocation method and system based on triangular fuzzy number weights provided in this invention has the following beneficial effects:

[0045] 1. This invention uses triangular fuzzy numbers to characterize path evaluation attribute values, which can more realistically depict the fuzziness and uncertainty in the traffic environment and overcome the shortcomings of traditional precise numerical models in describing the situation. By combining deviation maximization with moisture value analysis, a comprehensive attribute weight is obtained through weighted fusion. The dual consideration mechanism not only ensures the distinguishability between attributes but also reduces the decision-making risk caused by uncertainty, making the weight allocation more scientific and reliable.

[0046] 2. This invention employs the K-Means algorithm for path search. By selecting a representative subset of vertices for pre-computation and combining it with an incremental update strategy, the shortest path subtree affected by changes in road network weights is updated locally only, avoiding the overhead of global recomputation. The dynamic travel time prediction function based on the traffic impedance model can accurately reflect the coupled effects of departure time, real-time traffic flow, and capacity, effectively predicting path travel time under occasional congestion. Through time discretization and dynamic loading of traffic flow, the model can more accurately depict the dynamic changes in road network traffic conditions, providing real-time and accurate basis for path allocation.

[0047] 3. This invention achieves a comprehensive balance between safety and efficiency indicators by dynamically adjusting the weights of lane-changing safety and comfort using a triangular fuzzy controller. Through the organic integration of triangular fuzzy number theory, dynamic path search technology, and multi-objective optimization methods, it effectively solves the shortcomings of traditional shortest path algorithms in dealing with traffic environment uncertainties, dynamics, and multi-objective optimization needs, providing a more scientific, efficient, and safe solution for path planning problems in intelligent transportation systems.

[0048] These or other aspects of the invention will become more apparent from the following description of embodiments. It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. In the drawings:

[0050] Figure 1 is a flowchart of a traffic dynamic shortest path allocation method based on triangular fuzzy number weights according to the present invention.

[0051] Figure 2 shows the critical value and cumulative frequency curves in the traffic dynamic shortest path allocation method based on triangular fuzzy number weights of the present invention.

[0052] Figure 3 is a flowchart of the traffic dynamic shortest path allocation model based on triangular fuzzy weights in the traffic dynamic shortest path allocation method of the present invention.

[0053] Figure 4 shows the traffic dynamic shortest path lane-changing diagram in a traffic dynamic shortest path allocation method based on triangular fuzzy number weights according to the present invention.

[0054] Figure 5 is a schematic diagram of the state of the graph when vertex s1 is closed in a traffic dynamic shortest path allocation method based on triangular fuzzy number weights according to the present invention.

[0055] Figure 6 is a comparison of the running times of Hraf and K-means in a traffic dynamic shortest path allocation method based on triangular fuzzy number weights according to the present invention.

[0056] Figure 7 is a performance comparison chart of three full-source shortest path algorithms in a traffic dynamic shortest path allocation method based on triangular fuzzy number weights according to the present invention. Detailed Implementation

[0057] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings, but it should be understood that the scope of protection of the present invention is not limited to the specific embodiments.

[0058] Unless otherwise expressly stated, throughout the specification and claims, the term "comprising" or its variations such as "including" or "comprises" shall be understood to include the stated elements or components without excluding other elements or other components.

[0059] As shown in Figure 1, one embodiment of the present invention provides a method for dynamic shortest path allocation based on triangular fuzzy number weights, including the following steps:

[0060] Step S10: Obtain the evaluation attribute values ​​of candidate paths in the traffic network, represent each attribute value as a triangular fuzzy number, and construct a multi-attribute decision matrix based on triangular fuzzy numbers;

[0061] Step S20: Based on the multi-attribute decision matrix, calculate the discrimination weight of each attribute value in the evaluation result by using deviation maximization, combine the entropy value of all path evaluation attribute values ​​under each attribute value, calculate the uncertainty weight of attribute information, and weight and fuse the attribute weight based on discrimination and the uncertainty weight of attribute information to obtain the comprehensive attribute weight.

[0062] Step S30: Based on the comprehensive attribute weights, select a representative subset of vertices from the road network vertex set to pre-calculate the shortest path distance between all vertex pairs, and quickly retrieve the shortest path set;

[0063] Step S40: Discretize time using the point of sudden change in road traffic state as the node, dynamically load traffic flow based on traffic impedance model, establish a travel time prediction function related to vehicle departure time, real-time road segment flow and road segment capacity, and obtain the predicted travel time of each path at different departure times.

[0064] Step S50: Based on the comprehensive attribute weights, the shortest path set, and the predicted travel time, a comprehensive evaluation of the path is performed, the optimal recommended path is assigned to the vehicle, and the path scheme and travel time information are output.

[0065] The traffic dynamic shortest path allocation method in this embodiment reflects the role of various indicator factors in the traffic dynamic shortest path allocation process, directly affecting the result. Based on the triangular fuzzy number weighting perspective, this invention provides fuzzy information, triangular fuzzy numbers, and their sorting methods, making it an effective method for analyzing the weights of indicator factors with fuzzy information.

[0066] In this embodiment, the triangular fuzzy number is represented as: ,in, The left endpoint of the fuzzy number, The center point of the fuzzy number, Let be the right endpoint of the fuzzy number; the membership function of the triangular fuzzy number is defined as a piecewise linear function, used to characterize the fuzziness of attribute values, and the membership function is:

[0067] ;

[0068] In the formula, Let be any candidate value within the universe of discourse; when or At that time, the membership degree is 0.

[0069] In this embodiment, the judgment matrix is ​​determined by establishing membership functions. A set of membership functions is established using cumulative frequency curves. As shown in Figure 2, the cumulative frequency curve of the traffic dynamic shortest path allocation index is used. If the proposed situation is categorized into five levels: good, relatively good, moderate, relatively poor, and poor, then the index values ​​corresponding to the 15%, 30%, 50%, 65%, and 80% cumulative frequency values ​​on the cumulative frequency curve can be selected as critical values, thus obtaining the membership functions shown in Table 1.

[0070] Table 1 Membership Function Table

[0071]

[0072] For m indicators under a certain criterion C, the membership degree r of each indicator i to level j can be obtained from the corresponding membership function, thus yielding an evaluation matrix R:

[0073] .

[0074] Fuzzy weighting handles uncertainties influenced by numerous factors, such as the ambiguity of language and human intervention. Other factors also affect a driver's route selection, including travel time, travel distance, number of traffic signals, recreational routes, and route difficulty. While travel time and distance can be used to find the shortest path, the optimal route is one that simultaneously considers multiple factors, including traffic signal count, recreational routes, and route difficulty. Therefore, fuzzy theory can be used to derive the optimal route that comprehensively considers various factors, as shown in Figure 3.

[0075] In this embodiment, the calculation of the discrimination weight specifically includes: measuring the difference in evaluation results of each attribute through the deviation maximization model, and the greater the difference, the higher the weight; the calculation of the uncertainty weight specifically includes: measuring the uncertainty of evaluation information under each attribute through the entropy value, and the higher the uncertainty, the lower the weight adjustment factor; the weighted fusion adopts a linear weighting method, and the fusion coefficient is determined based on the reliability index.

[0076] In actual dynamic shortest path selection, under different working conditions, lane-changing trajectories with fixed weight coefficients are difficult to adapt to the selection of the optimal lane-changing trajectory under multiple working conditions. The inputs of the triangular fuzzy controller are the lane-changing safety factor UL, the vehicle speed, and the road surface adhesion coefficient, and the output is the lane-changing comfort weight.

[0077] According to relevant regulations, the longitudinal safety distance is the responsibility of the following vehicle. Therefore, the lane-changing space for vehicle M is jointly determined by vehicles F0, F1, and R1. Let M be the initial longitudinal distance between vehicle M and vehicle F0; The minimum safe lane-changing distance between vehicle M and vehicle F0 is defined as the ratio of the lane-changing space coefficient between M and F.

[0078] ;

[0079] After the lane change is completed, M, F1 and R1 must maintain a safe distance, let S be the safe distance; The initial longitudinal distance between the two vehicles; To define the minimum safe lane-changing distance between vehicle M and vehicle F1, the lane-changing space coefficient between M and F1 is defined as:

[0080] ;

[0081] Similarly, let The initial longitudinal distance between the two vehicles; To define the minimum safe lane-changing distance between vehicle M and vehicle R1, the lane-changing space coefficient between M and R1 is defined as follows:

[0082] ;

[0083] , , The smallest value reflects the most confined lane-changing space between the corresponding vehicle and the vehicle itself; therefore, the lane-changing space coefficient is defined. :

[0084] ;

[0085] The universe of discourse for lane-changing safety factor is divided into four fuzzy subsets {S, MS, M, B}. When the lane-changing time remains constant, as the vehicle speed increases, the comfort weight w should be reduced to improve lane-changing efficiency and safety. The vehicle speed universe of discourse is defined as B{0,1,2,3,4,5,6,7,8,9,10,11,12}, with a quantization factor K=0.1. The vehicle speed universe of discourse is divided into three fuzzy subsets {S, M, B}. When the road surface adhesion coefficient decreases, lateral acceleration should be minimized as much as possible while ensuring lane-changing safety. The lane-changing comfort weight w can be appropriately increased to improve lane-changing lateral safety. The universe of discourse for adhesion coefficient is defined as C{0,1,2,3,4,5,6,7,8,9}, with a quantization factor K=10. The universe of discourse for the adhesion coefficient is divided into three fuzzy subsets {S,M,B}, and the universe of discourse for the comfort weight is defined as D{0,1,2,3,4,5,6,7,8,9,10}, with a scaling factor K=0.1. The universe of discourse D is further divided into eleven fuzzy subsets {A,B,C,D,E,F,G,H,I,J,K}. Based on the above analysis, the triangular fuzzy control rules are formulated as shown in Table 2.

[0086] Table 2 Triangular Fuzzy Control Rules

[0087] When the adhesion coefficient is 0.3, the lateral acceleration threshold is 0.25s, and the corresponding shortest lane-changing time is 3.05s. Under this condition, the critical collision time corresponding to the extreme lane-changing trajectory is t=1.75s, the minimum lane-changing safety distance Ssafe1 is 17.3 m, and the lane-changing safety factor UL=UL1=2.35. At this time, the comfort weight output by the triangular fuzzy controller is 0.76, and the optimal lane-changing time is 6.24s. Under this condition, the lane-changing trajectories based on triangular fuzzy variable weight and fixed weight programming are shown in Figure 4.

[0088] The lane-changing trajectory based on triangular fuzzy variable weights corresponds to a critical collision time tc = 3.28s. From the lane-changing start time ts to time tc, the minimum longitudinal distance between vehicle M and vehicle F0 is 7.04 m; during the lane-changing process, the maximum lateral acceleration is 0.57 m / s², and the maximum impact is 0.96 m / s³. Therefore, the lane-changing trajectory planning strategy based on triangular fuzzy variable weights can prioritize lane-changing comfort under conditions of sufficient lane-changing space, low collision risk, but low road surface adhesion coefficient and high risk of sideslip. By adjusting the weight values ​​through triangular fuzzy control, compared with the fixed-weight lane-changing trajectory planning algorithm, the maximum lateral acceleration is reduced by 38.6%, and the maximum impact is reduced by 67.4%, thus effectively improving lateral safety and lane-changing comfort while ensuring no collision occurs.

[0089] In this embodiment, when pre-calculating the shortest path distance between all vertex pairs, K-Means is used for path search, including the following steps:

[0090] Preprocessing stage: Select a subset R from the vertex set V as representative vertices, and pre-calculate the shortest path distance between all pairs of vertices in R;

[0091] Dynamic retrieval phase: When the weights of the road network edges change, the set of affected vertices is identified, and an incremental update strategy is used to locally correct the pre-calculated data;

[0092] Path generation stage: Combining the A* algorithm and candidate path management mechanism, K shortest paths are quickly retrieved using pre-calculated distances.

[0093] This invention employs the K-Means algorithm for path search, pre-compiling a representative subset of vertices and combining it with an incremental update strategy. It only locally updates the shortest path subtree affected by changes in road network weights, avoiding the overhead of global recomputation. The dynamic travel time prediction function based on the traffic impedance model accurately reflects the coupled effects of departure time, real-time traffic flow, and capacity, effectively predicting path travel time under occasional congestion. Through time discretization and dynamic loading of traffic flow, the model can more accurately depict the dynamic changes in road network traffic conditions, providing real-time and accurate data for path allocation.

[0094] This embodiment proposes using the K-Means algorithm for path search. First, the A* algorithm is applied until the target node t is found, then A* is paused. At this point, the first path st has been found. Next, the following iterative strategy is used to find the remaining k-1 shortest paths. The process is as follows: traverse all vertices on the i-th shortest path. Select the shortest path from the candidate path set as the (i+1)-th shortest path. Recursively repeat the above process until k shortest paths are found.

[0095] This embodiment features on-the-fly search and can be guided by heuristic functions. When recursively solving for k shortest paths, the A* algorithm can be restarted or paused as needed. The heuristic search characteristic is mainly reflected in the implementation of A*.

[0096] At the start of the iteration, the candidate path list C is empty. Candidate paths are obtained by traversing the first shortest path, specifically as follows: Starting from the last vertex S6 of the first shortest path, traverse along the shortest path to vertex S2, which requires visiting vertices S6, S4, and S2. For vertex S6, there are three edges; two edges (S1, S6) and (S4, S6) are side trackedges. These two edges can represent two candidate paths, which are added to the candidate path queue. When processing edge {S1, s6}, since vertex s1 is not closed in A*, A* needs to be restarted until s1 closes. Figure 5 shows the graph state when s1 closes. When s1 closes, d(s1) is calculated, and the length of the candidate path {(s1, s6)} is w(s1s6) + d(s1) = 10. This candidate path is then added to the candidate path queue C. Table 3 shows the results of the first shortest path traversal.

[0097] Table 3 Iteration results of the first shortest path

[0098]

[0099] Table 4 shows the final 7 paths obtained.

[0100]

[0101] Repeat the above iterative steps until all 7 shortest paths are found. Table 4 lists the 7 shortest paths found at the end. The st path is the final actual path.

[0102] Figure 6 shows a comparison of the running time of the Hraf algorithm and the K-means algorithm. It can be seen from the figure that K has a much greater impact on Hraf than on K-means. This is because Hraf adopts an iterative strategy. Starting from the first path, the computational cost of calculating k paths is almost proportional to k. In contrast, K-means first builds a complex path graph and then finds k paths in that graph. Once the path graph is built, the workload of finding k paths is very small, so the time curve of K-means is very flat.

[0103] In this embodiment, the incremental update strategy is as follows: when a change in edge weight is detected, only the shortest path subtree affected by the edge is recalculated, and the relevant candidate path set is updated to avoid recalculating the entire path network; the strategy maintains the set of affected vertices. To achieve, among which, Representing an edge The subset consisting of the endpoint and all its descendant nodes.

[0104] When the weights of some traffic edges change, it is not necessary to recalculate the entire set of access points from scratch; instead, the recalculation starts from the nodes that are affected. The pseudocode for updating access nodes is shown in Table 5. Em represents the set of edges whose weights have changed; V represents the set of vertices whose outgoing edges in the cover graph at layer t have changed; R represents the set of vertices that the shortest path tree depends on.

[0105] Table 5 Update Algorithm for Access Nodes

[0106]

[0107] First, each vertex in the tree is associated with two attributes: the distance to the root and the corresponding parent node, denoted as d(v,T) and p(v,T) respectively. During the algorithm execution, T gradually changes, and after the algorithm completes, it represents the correct shortest path tree. For a directed edge e, s(e) represents the starting point of the edge, E(e) represents the ending point, and w(e) represents the weight. I(v) represents the set of incoming edges, O(v) represents the set of outgoing edges, and B(v,T) represents the subset of vertices in tree T consisting of v itself and some of its descendant vertices. Specifically, in this embodiment, the notation Bmax(E(e),T) is used to represent the subset consisting of the ending point of edge e and all its descendant nodes.

[0108] The pseudocode for the shortest path tree update algorithm is shown in Table 6. We discuss two cases: increasing edge weights and decreasing edge weights. Note that if edge e is deleted, it can be seen as w(e) increasing to infinity, a special case of the first scenario. In the first case (increasing edge weights), this embodiment first determines whether edge e is in the existing shortest path tree. If not, the algorithm terminates. Otherwise, we denote the set of the endpoint of edge e and all its descendant vertices as N, noting that N already covers all affected vertices.

[0109] Table 6. Shortest Path Tree Update Algorithm

[0110]

[0111] Experimental environment: Intel Xeon E3-1220 server, CPU frequency 3.1GHz, memory 32GB. Linux 64-bit operating system, compiled with 02 optimization options. This embodiment implements and compares three full-source shortest path algorithms, running these three algorithms on a real road network dataset. Table 7 lists the running time of the three algorithms and the average number of vertices expanded when calculating single-source shortest paths. Figure 7 is selected. The experimental results show that the algorithm in this application is 8.5 to 13.9 times faster than the nDIJ algorithm, a performance improvement of one order of magnitude, and 3 to 5.4 times faster than the nBFS algorithm. On average, it only expands 18.8% to 37.5% of the vertices.

[0112] Table 7. Performance Comparison of Three All-Source Shortest Path Algorithms

[0113]

[0114] In this embodiment, the travel time prediction function is: ,in, The vehicle's departure time. For real-time traffic flow, For the traffic capacity of the road segment; function It is obtained by fitting historical data or training by machine learning, and is used to reflect the dynamic changes in road resistance under occasional congestion; dynamic loading is based on a time discretization strategy, which divides continuous time into discrete intervals for processing.

[0115] In this embodiment, the path is comprehensively evaluated based on the comprehensive attribute weights, the shortest path set, and the predicted travel time. When assigning the optimal recommended path to the vehicle, a multi-objective decision-making method is used. The multi-objective decision-making includes at least one of the following optimization objectives: minimizing total travel time, minimizing path distance, equalizing traffic flow, and optimizing lane-changing safety and comfort. The multi-objective decision-making uses a triangular fuzzy controller to dynamically adjust the weights of each objective. The controller's inputs include the lane-changing safety coefficient, the vehicle speed, and the road surface adhesion coefficient, and the output is the comfort weight.

[0116] In this embodiment, the triangular fuzzy controller includes:

[0117] The universe of discourse for lane-changing safety factor, vehicle speed, and adhesion coefficient is divided into fuzzy subsets, and a fuzzy control rule table is formulated.

[0118] The comfort weights are output by defuzzification and used to adjust the weights of comfort indicators in path evaluation.

[0119] This invention uses triangular fuzzy numbers to characterize path evaluation attribute values, which can more realistically depict the fuzziness and uncertainty in the traffic environment and overcome the shortcomings of traditional precise numerical models in describing the situation. By combining deviation maximization with moisture value analysis, a comprehensive attribute weight is obtained through weighted fusion. This dual-consideration mechanism ensures the distinguishability between attributes while reducing the decision-making risk caused by uncertainty, making the weight allocation more scientific and reliable.

[0120] Another embodiment of the present invention provides a traffic dynamic shortest path allocation system based on triangular fuzzy number weights, comprising:

[0121] The data acquisition module is used to obtain the topology of the traffic network, multiple evaluation attribute values ​​of each candidate path, real-time traffic flow data, and vehicle status information.

[0122] The weight calculation module is communicatively connected to the data acquisition module. It is used to represent each evaluation attribute value as a triangular fuzzy number to construct a multi-attribute decision matrix. Based on the decision matrix, it calculates the discrimination weight of each attribute using the deviation maximization method. It also calculates the uncertainty weight of each attribute using the entropy value and merges the two types of weights to obtain the comprehensive attribute weight.

[0123] The path search module is communicatively connected to the weight calculation module and is used to perform path search using the K-Means algorithm based on the comprehensive attribute weights. The path search includes: selecting a representative subset of vertices from the network vertex set and pre-calculating the shortest path distance between all vertex pairs; and using an incremental update strategy to locally correct the pre-calculated data when the weights of the network edges change, so as to quickly retrieve the top K shortest paths.

[0124] The travel time prediction module is communicatively connected to the data acquisition module. It is used to discretize time with the point of sudden change in road traffic state as the node, dynamically load traffic flow based on traffic impedance model, and model the travel time of road segment as a function related to vehicle departure time, real-time road segment flow and road segment capacity, so as to predict the dynamic travel time of each path.

[0125] The decision and output module is communicatively connected to the weight calculation module, the path search module, and the travel time prediction module, respectively. It is used to combine the comprehensive attribute weights, the set of K shortest paths, and the predicted dynamic travel time to allocate the optimal path to the vehicle through a multi-objective decision-making method, and output the path scheme and the estimated travel time information.

[0126] In this embodiment, the triangular fuzzy number used to represent the evaluation attribute value in the weight calculation module is: ,in, The left endpoint of the fuzzy number, The center point of the fuzzy number, Let be the right endpoint of the fuzzy number; the membership function of the triangular fuzzy number is defined as a piecewise linear function, used to characterize the fuzziness of attribute values, and the membership function is:

[0127] ;

[0128] In the formula, Let be any candidate value within the universe of discourse; when or At that time, the membership degree is 0.

[0129] In this embodiment, the path search module includes:

[0130] The preprocessing unit is used to select a subset R from the vertex set V as representative vertices and pre-calculate the shortest path distance between all pairs of vertices in R.

[0131] The dynamic processing unit is used to identify the set of affected vertices when the weights of the road network edges change, and to locally correct the pre-calculated data in the preprocessing unit using an incremental update strategy.

[0132] The path generation unit is used to combine the A* algorithm and the candidate path management mechanism to quickly retrieve K shortest paths using the pre-calculated distance.

[0133] This invention achieves a comprehensive balance between safety and efficiency indicators by dynamically adjusting the weights of lane-changing safety and comfort using a triangular fuzzy controller. Through the organic integration of triangular fuzzy number theory, dynamic path search technology, and multi-objective optimization methods, it effectively solves the shortcomings of traditional shortest path algorithms in dealing with traffic environment uncertainties, dynamism, and multi-objective optimization needs, providing a more scientific, efficient, and safe solution for path planning problems in intelligent transportation systems.

[0134] The foregoing description of specific exemplary embodiments of the invention is for illustrative and explanatory purposes. These descriptions are not intended to limit the invention to the precise forms disclosed, and it will be apparent that many changes and variations can be made in accordance with the foregoing teachings. The exemplary embodiments were chosen and described in order to explain the specific principles of the invention and its practical application, thereby enabling those skilled in the art to implement and utilize various different exemplary embodiments of the invention, as well as various different choices and variations. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A method for assigning dynamic shortest paths in traffic based on triangular fuzzy number weights, characterized in that, The process includes the following steps: obtaining the evaluation attribute values ​​of candidate paths in the traffic network, representing each attribute value as a triangular fuzzy number, and constructing a multi-attribute decision matrix based on the triangular fuzzy number; Based on the multi-attribute decision matrix, the discrimination weight of each attribute value in the evaluation result is calculated by using deviation maximization. The uncertainty weight of the attribute information is calculated by combining the entropy value of all path evaluation attribute values ​​under each attribute value. The attribute weight based on discrimination and the uncertainty weight of attribute information are then weighted and fused to obtain the comprehensive attribute weight. Based on the comprehensive attribute weights, a representative subset of vertices is selected from the road network vertex set to pre-calculate the shortest path distance between all vertex pairs, and the shortest path set is quickly retrieved. By discretizing time using the point of sudden change in road traffic conditions as nodes, and dynamically loading traffic flow based on the traffic impedance model, a travel time prediction function related to vehicle departure time, real-time road segment flow, and road segment capacity is established to obtain the predicted travel time of each path at different departure times. The path is comprehensively evaluated based on the combined attribute weights, the shortest path set, and the predicted travel time. The optimal recommended path is assigned to the vehicle, and the path scheme and travel time information are output.

2. The traffic dynamic shortest path allocation method based on triangular fuzzy number weights as described in claim 1, characterized in that, The triangular fuzzy number is represented as: ,in, The left endpoint of the fuzzy number, The center point of the fuzzy number, Let be the right endpoint of the fuzzy number; the membership function of the triangular fuzzy number is defined as a piecewise linear function, used to characterize the fuzziness of attribute values, and the membership function is: In the formula, Let be any candidate value within the universe of discourse; when or At that time, the membership degree is 0.

3. The traffic dynamic shortest path allocation method based on triangular fuzzy number weights as described in claim 2, characterized in that, The calculation of the discrimination weight specifically includes: measuring the difference in evaluation results of each attribute through the deviation maximization model, and the greater the difference, the higher the weight; the calculation of the uncertainty weight specifically includes: measuring the uncertainty of evaluation information under each attribute through the entropy value, and the higher the uncertainty, the lower the weight adjustment factor; the weighted fusion adopts a linear weighting method, and the fusion coefficient is determined based on the reliability index.

4. The traffic dynamic shortest path allocation method based on triangular fuzzy number weights as described in claim 3, characterized in that, When pre-calculating the shortest path distance between all vertex pairs, K-Means is used for path search, including the following steps: Preprocessing stage: Select a subset R from the vertex set V as representative vertices, and pre-calculate the shortest path distance between all vertex pairs in R; Dynamic retrieval stage: When the edge weights of the road network change, identify the set of affected vertices, and use an incremental update strategy to locally correct the pre-calculated data; Path generation stage: Combine the A* algorithm and the candidate path management mechanism to quickly retrieve K shortest paths using the pre-calculated distances.

5. The traffic dynamic shortest path allocation method based on triangular fuzzy number weights as described in claim 4, characterized in that, The incremental update strategy is as follows: when a change in edge weight is detected, only the shortest path subtree affected by the edge is recalculated, and the relevant candidate path set is updated to avoid recalculating the entire path network; this strategy maintains the set of affected vertices. To achieve, among which, Representing an edge The subset consisting of the endpoint and all its descendant nodes.

6. The traffic dynamic shortest path allocation method based on triangular fuzzy number weights as described in claim 5, characterized in that, The travel time prediction function is: ,in, The vehicle's departure time. For real-time traffic flow, For the traffic capacity of the road segment; function It is obtained by fitting historical data or training by machine learning, and is used to reflect the dynamic changes in road resistance under occasional congestion; dynamic loading is based on a time discretization strategy, which divides continuous time into discrete intervals for processing.

7. The traffic dynamic shortest path allocation method based on triangular fuzzy number weights as described in claim 6, characterized in that, The path is comprehensively evaluated based on the comprehensive attribute weights, the shortest path set, and the predicted travel time. When assigning the optimal recommended path to the vehicle, a multi-objective decision-making method is used. The multi-objective decision-making includes at least one of the following optimization objectives: minimizing the total travel time, minimizing the path distance, equalizing traffic flow, and optimizing lane-changing safety and comfort. The multi-objective decision-making uses a triangular fuzzy controller to dynamically adjust the weights of each objective. The controller's inputs include the lane-changing safety coefficient, the vehicle speed, and the road surface adhesion coefficient, and the output is the comfort weight.

8. The traffic dynamic shortest path allocation method based on triangular fuzzy number weights as described in claim 7, characterized in that, The triangular fuzzy controller includes: dividing the universe of discourse of lane-changing safety factor, vehicle speed and adhesion coefficient into fuzzy subsets and formulating a fuzzy control rule table; and outputting comfort weights through defuzzification to adjust the comfort index weights in path evaluation.

9. A traffic dynamic shortest path allocation system based on triangular fuzzy number weights, characterized in that, The system is used to perform the steps of the traffic dynamic shortest path allocation method based on triangular fuzzy number weights as described in any one of claims 1-8. The traffic dynamic shortest path allocation system includes: a data acquisition module for acquiring the topology of the traffic network, multiple evaluation attribute values ​​of each candidate path, real-time traffic flow data, and vehicle status information; a weight calculation module, communicatively connected to the data acquisition module, for representing each evaluation attribute value as a triangular fuzzy number to construct a multi-attribute decision matrix, and calculating the discrimination weight of each attribute using the deviation maximization method based on the decision matrix, and combining the calculation of the uncertainty weight of each attribute using the entropy value, and fusing the two types of weights to obtain a comprehensive attribute weight; and a path search module, communicatively connected to the weight calculation module, for performing path search using the K-Means algorithm based on the comprehensive attribute weights, including: selecting representative values ​​from the set of network vertices. The system includes a vertex subset and pre-calculates the shortest path distance between all vertex pairs. It also employs an incremental update strategy to locally correct the pre-calculated data when the weights of the road network edges change, enabling rapid retrieval of the top K shortest paths. A travel time prediction module, communicatively connected to the data acquisition module, discretizes time using abrupt changes in road traffic conditions as nodes. Based on a traffic impedance model, it dynamically loads traffic flow, modeling road segment travel time as a function related to vehicle departure time, real-time road segment flow, and road segment capacity to predict the dynamic travel time of each path. A decision and output module, communicatively connected to the weight calculation module, the path search module, and the travel time prediction module, combines the comprehensive attribute weights, the set of K shortest paths, and the predicted dynamic travel time to allocate the optimal path to vehicles using a multi-objective decision-making method, and outputs the path scheme and estimated travel time information.

10. The traffic dynamic shortest path allocation system based on triangular fuzzy number weights as described in claim 9, characterized in that, The path search module includes: a preprocessing unit, used to select a subset R from the vertex set V as representative vertices and pre-calculate the shortest path distance between all vertex pairs in R; a dynamic processing unit, used to identify the set of affected vertices when the edge weights of the road network change, and use an incremental update strategy to locally correct the pre-calculated data in the preprocessing unit; and a path generation unit, used to combine the A* algorithm and the candidate path management mechanism to quickly retrieve K shortest paths using the pre-calculated distances.

Images