A truck route planning method and system fusing dynamic traffic flow and differentiated charging policy

Abstract

The application provides a truck path planning method and system fusing dynamic traffic flow and differentiated charging policy, relates to the technical field of truck transportation path planning, and first receives a roadside sensing data set of a target road network in a planning time interval, then carries out time-varying impedance assignment on a road section unit of the road network by using the roadside sensing data set, generates a road network impedance distribution state diagram, projects and forms a time-varying mixed cost road network representation structure by aligning charging policy configuration information, then performs time-varying path search on the structure according to a given truck trip starting point and ending point, outputs a complete path section sequence and the driving-in and driving-out time of each section unit, and finally sends truck path planning scheduling information to a target scheduling terminal, so that accurate guidance is provided for truck driving, and the reliability and accuracy of truck transportation are improved.

Description

Technical Field

[0001] This invention relates to the field of truck transportation route planning technology, and more specifically, to a truck route planning method and system that integrates dynamic traffic flow and differentiated pricing policies. Background Technology

[0002] In the truck transportation sector, route planning is a crucial step in improving transportation efficiency and reducing costs. Traditional truck route planning methods often only consider static information about the road network, such as fixed road lengths and traffic restrictions, while ignoring the impact of dynamic changes in traffic flow on truck driving. However, actual road traffic conditions are constantly changing, with significant differences in traffic volume and speed across different road segments at different times. Routes planned solely based on static information may consume a lot of time during actual travel due to traffic congestion, failing to meet the demands of efficient transportation.

[0003] Meanwhile, most existing route planning methods do not fully consider the impact of differentiated toll policies on truck route selection. Different road segments have different truck toll coefficients at different times, and truck drivers tend to choose routes with lower tolls to reduce transportation costs. However, existing planning methods fail to effectively integrate the dynamic changes in traffic flow with differentiated toll policies, resulting in planned routes that either cannot adapt to real-time traffic conditions or cannot effectively reduce transportation costs, thus failing to achieve a dual optimization of transportation efficiency and cost. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a truck route planning method integrating dynamic traffic flow and differentiated toll policies, the method comprising: The system receives roadside sensor datasets and toll policy configuration sets for the target road network within the planned time interval. The roadside sensor datasets include average traffic flow rate records and traffic flow space occupancy records for each road segment unit at multiple collection times. The toll policy configuration sets include truck toll coefficients and toll coefficient switching time sequences for each road segment unit within multiple time intervals. The average traffic flow rate record and the traffic flow space occupancy rate record of the cross section are used to perform time-varying impedance assignment processing on all road segment units of the target road network to generate a road network impedance distribution state map that is dynamically updated with time segments. The road segment impedance value in the road network impedance distribution state map is determined by fusing the traffic flow state mapping relationship of the average traffic flow rate record and the traffic flow space occupancy rate record of the cross section. The truck toll coefficients in the toll strategy configuration set are projected onto the road network impedance distribution state map in a time dimension alignment process, so that each road segment unit carries the road segment impedance value and the corresponding truck toll coefficient at each analysis step in the time series, forming a time-varying hybrid cost road network representation structure. Based on the given starting and ending points of the truck trip, a time-varying path search process is performed on the time-varying hybrid cost road network representation structure. The time-varying path search process reads the road segment impedance value and truck passage cost coefficient corresponding to the entry time of the road segment when traversing adjacent road segment units, and combines them to calculate a comprehensive path selection cost. The complete path segment sequence and the entry and exit times of each road segment unit in the path segment sequence are output. The complete route segment sequence, the entry time and the departure time of each segment unit are combined to form truck route planning and scheduling information, and the truck route planning and scheduling information is sent to the target scheduling terminal to trigger truck driving route guidance operation.

[0005] Furthermore, embodiments of the present invention also provide a truck route planning system that integrates dynamic traffic flow and differentiated toll policies, comprising: A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the aforementioned truck routing method integrating dynamic traffic flow and differentiated tolling policies by executing the machine-executable instructions.

[0006] Based on the above, by receiving roadside sensor datasets and toll policy configuration sets of the target road network within the planning time interval, dynamic information on road network traffic flow and information on differentiated toll policies were obtained. Time-varying impedance values ​​were assigned to all road network segments using cross-sectional average traffic flow rate records and cross-sectional traffic flow space occupancy records, generating a road network impedance distribution state map that dynamically updates with time segments. This accurately reflects the ease or difficulty of passage for each road segment in different time periods. The truck toll fee coefficients from the toll policy configuration set were then time-aligned and projected onto the road network impedance distribution state map, forming a time-varying hybrid cost road network representation structure that comprehensively considers the impact of dynamic traffic flow changes and differentiated toll policies on route selection. In time-varying path search processing, the road segment impedance value and truck passage cost coefficient corresponding to the entry time of the road segment are read and combined to calculate the comprehensive path selection cost. This allows the planned path to adapt to real-time traffic conditions and effectively reduce transportation costs, achieving overall optimization of transportation efficiency and cost. The output complete path segment sequence and the entry and exit time information of each road segment unit provide accurate guidance for truck driving, improving the reliability and accuracy of truck transportation. Attached Figure Description

[0007] Figure 1 This is a schematic diagram illustrating the execution flow of the truck route planning method that integrates dynamic traffic flow and differentiated toll policies provided in this embodiment of the invention.

[0008] Figure 2 This is a schematic diagram of exemplary hardware and software components of a truck route planning system that integrates dynamic traffic flow and differentiated toll policies, provided in an embodiment of the present invention. Detailed Implementation

[0009] Figure 1 This is a flowchart illustrating a truck route planning method that integrates dynamic traffic flow and differentiated toll policies, provided in one embodiment of the present invention. A detailed description follows.

[0010] The truck route planning method integrating dynamic traffic flow and differentiated toll policies provided in this application can be applied to urban freight dispatching scenarios. In this scenario, the target road network covers urban expressways, arterial roads, and secondary arterial roads. The traffic status of each road segment changes over time, and different road segments implement differentiated truck toll policies at different times. After receiving roadside sensor data and toll policy configuration, this method dynamically generates route planning results to guide truck travel, balancing travel time and toll costs. It should be noted that the roadside sensor data collection only involves anonymized macroscopic traffic flow statistics and does not include vehicle license plate numbers or driver identification information. The data is protected during transmission and storage through encrypted channels and access control lists, complying with relevant data security laws and regulations.

[0011] Step S110: Receive the roadside sensor dataset and toll policy configuration set of the target road network within the planned time interval. The roadside sensor dataset contains the average cross-sectional traffic flow rate records and cross-sectional traffic flow space occupancy records of each road segment unit at multiple collection times. The toll policy configuration set contains the truck toll coefficient and toll coefficient switching time sequence of each road segment unit within multiple time intervals.

[0012] The roadside sensor dataset is received from roadside sensor devices deployed at various sections of the target road network. It is a set of structured data records. Each record contains a road segment unit identifier R, a collection time t, a rate value v, and an occupancy value o, where t is arranged at a fixed sampling interval. The toll policy configuration set is obtained from the rate management interface, using R as the index key. Each configuration record contains multiple time intervals for that road segment unit within the planned time interval, the corresponding cost coefficient c for each time interval, and the switching time sequence Q between adjacent time intervals, where Q is an ordered list of timestamps. v is dimensionless after being normalized by a preset rate normalization parameter, o is dimensionless after being normalized by a preset occupancy normalization parameter, and c is dimensionless after being normalized by a preset cost normalization parameter. The dimensionless values ​​v, o, and c are all dimensionless, and the dimensions on both sides of the equation are consistent.

[0013] Step S120: Use the cross-sectional average traffic flow rate record and the cross-sectional traffic flow space occupancy record to perform time-varying impedance assignment processing on all road segment units of the target road network, and generate a road network impedance distribution state map that is dynamically updated with time segments. The road segment impedance value in the road network impedance distribution state map is determined by the traffic flow state mapping relationship of the cross-sectional average traffic flow rate record and the cross-sectional traffic flow space occupancy record.

[0014] Step S121: Extract the average speed record of cross-sectional traffic flow of the target road segment unit in the roadside sensor data at the continuous acquisition time, and generate the average speed time series.

[0015] Using R as the retrieval key, all data records corresponding to the target road segment unit are filtered out from the roadside sensor dataset and sorted in ascending order by t. v is extracted sequentially from the sorted data records and arranged into a one-dimensional array V, which is the average rate time series. Each element in V corresponds to the rate value at a collection time.

[0016] Step S122: Extract the cross-sectional traffic flow space occupancy records of the target road segment unit in the roadside sensor dataset at continuous acquisition time, and generate a space occupancy time series.

[0017] Extract 'o' sequentially from the data records after ascending order in the same way, and arrange them into a one-dimensional array O. O is the space occupancy time series. O and V have the same length and the same index position corresponds to the same collection time.

[0018] Step S123: Input the speed sampling value in the average speed time series and the occupancy sampling value in the space occupancy time series at the same collection time into the traffic flow state mapping relationship, calculate the road segment impedance value of the target road segment unit at that collection time. The traffic flow state mapping relationship outputs an increased road segment impedance value when the occupancy sampling value increases and the speed sampling value decreases, and outputs a decreased road segment impedance value when the occupancy sampling value decreases and the speed sampling value increases.

[0019] For each data acquisition moment, the value of v at the current moment is extracted from V, and the value of o at the current moment is extracted from O. v and o are input into the traffic flow state mapping function f for calculation. f is monotonically increasing in regions where o increases and v decreases, and monotonically decreasing in regions where o decreases and v increases. The calculation method is as follows: Set a reference impedance value z0, a reference rate value vf, and a reference occupancy value of, and let z = z0 × (1 + α × (o / of)^β) × (vf / v)^γ, where α is the occupancy sensitivity coefficient, β is the occupancy nonlinearity exponent, and γ is the rate sensitivity exponent. For two adjacent data acquisition moments, if o increases and v decreases, then z at the second moment is greater than z at the first moment; if o decreases and v increases, then z at the second moment is less than z at the first moment.

[0020] Step S124: Traverse all road segment units included in the target road network, calculate the road segment impedance value of each road segment unit at each collection time according to the traffic flow state mapping relationship, and form a road segment unit impedance time series set. Each element in the road segment unit impedance time series set uniquely corresponds to the road segment impedance value of a road segment unit at a collection time.

[0021] For each road segment unit Ri in the target road network, the processing flow of steps S121 to S123 is executed to obtain the road segment impedance value zij of Ri at each acquisition time tj. All Ri and all zij corresponding to tj are summarized to form the road segment unit impedance time series set Zset. Each element in Zset is identified by a triple (Ri, tj, zij).

[0022] Step S125: Organize the road segment impedance values ​​in the road segment unit impedance time series set according to the spatial topological connection relationship of the road segment units. Based on the node adjacency table structure of the target road network, construct a road network impedance distribution state map with road segment units as the basic expression units. At each acquisition time, the road network impedance distribution state map generates an impedance snapshot layer that is isomorphic to the topological structure of the target road network.

[0023] The topology of the target road network is stored in the form of a node adjacency list. Each node represents a road intersection, and each edge represents a road segment unit. The node adjacency list stores the corresponding road segment unit identifier for each edge. For each data acquisition time tj, the impedance values ​​of all road segment units at time tj are retrieved from the Zset, and each impedance value is filled into the impedance attribute field of the corresponding edge in the node adjacency list. The completed node adjacency list is the impedance snapshot layer at time tj. Each impedance snapshot layer is completely isomorphic to the topology of the target road network. The impedance snapshot layers at all data acquisition times are arranged in chronological order to form a road network impedance distribution state map.

[0024] Step S126: Impedance transition filling is performed on the time gap between adjacent acquisition times. The road segment impedance value between adjacent acquisition times is generated by using the road segment impedance value of the previous acquisition time and the road segment impedance value of the next acquisition time through a time linear transition method, so that the road network impedance distribution state map forms an impedance change flow that is continuously and dynamically updated with time segments within the planned time interval.

[0025] For each pair of adjacent acquisition times tj and t(j+1), at each transition time tk between these two times, for each road segment unit Ri, take the impedance value zij of Ri at time tj and the impedance value zi(j+1) at time t(j+1), and calculate the transition time impedance value zik = zij + (zi(j+1) - zij) × (tk - tj) / (t(j+1) - tj). The impedance values ​​of all road segment units at all transition times constitute the impedance change flow, which continuously covers the planned time interval in the time dimension.

[0026] Step S127: Associate and store the transition impedance value of each road segment unit in the impedance change flow at different time segments with the corresponding road segment unit identifier and timestamp mark to obtain a time-refined road network impedance representation carrying the spatial attribute information and time attribute information of the road segment unit. Integrate the transition impedance values ​​of all road segment units belonging to the same time segment in the time-refined road network impedance representation into a time segment impedance distribution surface. The time segment impedance distribution surfaces of multiple consecutive time segments together form a road network impedance distribution state map that is dynamically updated with time segments.

[0027] Data from the impedance change stream is stored as triples (Ri, tk, zik), and this set of triples constitutes the time-refined road network impedance representation. Using a preset time segment length as the grouping window, records whose timestamps fall within the same time segment are grouped together. Within each group, the (Ri, zik) values ​​from all triples form the time segment impedance distribution surface for that time segment. The time segment impedance distribution surfaces of all time segments are arranged chronologically to form a dynamically updated road network impedance distribution state diagram that updates with each time segment.

[0028] Step S130: The truck toll fee coefficients in the toll strategy configuration set are projected onto the road network impedance distribution state map in a time dimension alignment process, so that each road segment unit carries the road segment impedance value and the corresponding truck toll fee coefficient at each analysis step in the time series, forming a time-varying hybrid cost road network representation structure.

[0029] Step S131: Read the freight toll coefficient sequence and the toll coefficient switching time sequence associated with the freight toll coefficient sequence corresponding to the specified road segment unit in the target road network from the toll strategy configuration set. The time interval boundary between two adjacent freight toll coefficients in the freight toll coefficient sequence is defined by the switching time mark in the toll coefficient switching time sequence.

[0030] Using the specified road segment unit R as the search key, the time period interval list and cost coefficient sequence for that road segment unit are read from the toll policy configuration set. The k-th element ck in the cost coefficient sequence corresponds to the k-th time period interval. Simultaneously, the switching time sequence Q is read, where the k-th element represents the boundary time between the k-th and (k+1)-th time period intervals. The values ​​of adjacent elements ck and c(k+1) in the cost coefficient sequence are different, separated by the switching time in Q.

[0031] Step S132: Construct the toll period boundary vector of the specified road segment unit based on the switching time marker in the toll factor switching time sequence. Each component of the toll period boundary vector represents the start and end time of a time interval.

[0032] Extract all switching times from Q. Use the start time of the planned time interval as the start time of the first time interval, and the first element in Q as the end time of the first time interval and the start time of the second time interval, and so on, until the end time of the planned time interval is used as the end time of the last time interval. Construct a charging time interval boundary vector B, where each element in B is a tuple (start time, end time).

[0033] Step S133: Extract the time-refined road network impedance representation corresponding to the specified road segment unit from the road network impedance distribution state map. The time-refined road network impedance representation contains a sequence of impedance values ​​indexed by timestamps. Each impedance value in the impedance value sequence corresponds to a collection time or transition time.

[0034] Using the R of a specified road segment unit as the search key, extract all triples (R, t, z) corresponding to that road segment unit from the time-refined road network impedance representation, arrange them in ascending order by t, and obtain the impedance value sequence Zseq indexed by timestamp.

[0035] Step S134: Starting from the beginning of the planned time interval, sample time points at a preset analysis step interval to generate a full-interval analysis time sequence. The time interval between two adjacent analysis times in the full-interval analysis time sequence is the analysis step interval.

[0036] Starting from the beginning time T0 of the planned time interval, the analysis time sequence A=[T0, T0+Δ, T0+2Δ, ..., T0+nΔ] is generated sequentially with the analysis step size interval Δ as the step size, where T0+nΔ is less than or equal to the end time of the planned time interval.

[0037] Step S135: For each analysis time in the full-interval analysis time sequence, extract the road segment impedance value corresponding to the timestamp of the analysis time from the impedance value sequence. At the same time, determine the associated truck toll fee coefficient according to the time interval to which the analysis time belongs in the toll period boundary vector. Pair the road segment impedance value and the truck toll fee coefficient to form the road segment mixed cost entry for the analysis time.

[0038] For each analysis time ak in A, the impedance value zk corresponding to the closest timestamp to ak is found in Zseq based on timestamp matching. At the same time, the time period interval into which ak falls is found in the toll period boundary vector B, and the cost coefficient ck corresponding to that time period interval is taken. Pairing zk and ck together to form (zk, ck) is the road segment mixed cost entry for time ak.

[0039] Step S136: Construct a mixed cost time series vector corresponding to the specified road segment unit. Each component position of the mixed cost time series vector stores a road segment mixed cost entry at the analysis time. The road segment mixed cost entry contains a road segment impedance value field and a truck passage cost coefficient field.

[0040] Arrange the road segment mixed cost entries of the specified road segment unit in chronological order at all analysis times to form the mixed cost time series vector of the road segment unit. The length of the vector is equal to the length of A, and the k-th component of the vector is (zk, ck) corresponding to the analysis time ak.

[0041] Step S137: Traverse all road segment units of the target road network, and for each road segment unit, perform the process of reading the freight vehicle toll coefficient sequence and toll coefficient switching time sequence from the toll policy configuration set, and generate their respective hybrid cost time series vectors to form a road network-level hybrid cost time series set. The hybrid cost time series vectors of each road segment unit in the road network-level hybrid cost time series set are associated and organized according to the spatial topological connection relationship of the road segment units. For adjacent road segment units with upstream and downstream connection relationships, the hybrid cost entries of adjacent road segment units at the same analysis time are aligned and arranged in the time-varying dimension to generate a time-varying hybrid cost road network representation structure.

[0042] Steps S131 to S136 are executed for each road segment unit in the target road network to obtain the mixed cost time series vector of each road segment unit. The mixed cost time series vectors of all road segment units constitute a road network-level mixed cost time series set. Using the node adjacency list as the skeleton, the mixed cost time series vector of each road segment unit is filled into the mixed cost attribute field of the corresponding edge. For adjacent road segment units with upstream and downstream connections, the mixed cost entries at the same analysis time are arranged in topological order in the edge structure of the adjacency list to form a time-varying mixed cost road network representation structure.

[0043] Step S140: Based on the given starting and ending points of the truck trip, perform time-varying path search processing on the time-varying hybrid cost road network representation structure. When traversing adjacent road segment units, the time-varying path search processing reads the road segment impedance value and truck passage cost coefficient corresponding to the entry time of the road segment and combines them to calculate the comprehensive path selection cost. Output the complete path segment sequence and the entry and exit times of each road segment unit in the path segment sequence.

[0044] Step S141: Map the starting position of the truck's journey to the nearest network node in the time-varying hybrid cost road network representation structure as the starting node for path search, and map the ending position of the truck's journey to the nearest network node in the time-varying hybrid cost road network representation structure as the ending node for path search.

[0045] Calculate the Euclidean distance between the starting point of the truck's journey and each node in the target road network, and select the node with the smallest Euclidean distance as the starting node Ns for the path search. Map the ending point of the truck's journey to the ending node Ne for the path search in the same way.

[0046] Step S142: Set the earliest entry time of the path search starting node within the planned time interval as the path search starting time point, and create a path tree with the path search starting node as the root node in the path search data structure. The root node of the path tree records the path search starting node identifier and the path search starting time point.

[0047] Let T0 be the starting time of the planning time interval, and use T0 as the starting time point for path search. Create a path tree data structure. The root node of the path tree stores Ns and T0. Each node in the path tree contains the current road segment unit identifier, entry time, parent node pointer, and cumulative path cost from the root node to the current node.

[0048] Step S143: Starting from the path search starting node, visit each neighbor node that is directly connected to the current node through a road segment unit in sequence. For the target neighbor node that is currently traversed, calculate the road segment entry time of the road segment unit traversed from the current node to the target neighbor node.

[0049] Let the current node be Nu, its arrival time be Tu, and the road segment from Nu to the target neighbor node Nv be Euv, with Euv having a geometric length parameter L. Calculate the departure time Tu from Nu based on its arrival time Tu; then the arrival time Tv of the target neighbor node Nv is equal to Tu.

[0050] Step S144: Based on the entry time of the road segment, find the road segment impedance value and truck toll coefficient of the corresponding road segment unit of the target neighbor node in the time-varying hybrid cost road network representation structure at the entry time of the road segment. Convert the found road segment impedance value into a time cost component through a preset time value coefficient, and use the truck toll coefficient as a monetary cost component. Input the time cost component and the monetary cost component into the comprehensive path selection cost fusion function. The comprehensive path selection cost fusion function performs a weighted sum of the time cost component and the monetary cost component to generate the comprehensive path selection cost of this visit, expressed in a unified monetary unit.

[0051] Using Tv as the time index, the segment impedance value z and cost factor c of segment unit Euv at time Tv are found in the time-varying hybrid cost road network representation structure. z is multiplied by a preset time value factor λ to obtain the time cost component C_time = z × λ. The cost factor c is multiplied by the segment unit geometric length parameter L to obtain the monetary cost component C_fee = c × L. The comprehensive path selection cost fusion function calculates the comprehensive path selection cost C_total = w1 × C_time + w2 × C_fee, where w1 and w2 are preset weighting coefficients, and C_total is in monetary units.

[0052] Step S145: The comprehensive path selection cost is added to the cumulative path cost in the path tree from the starting node of the path search to the target neighbor node, to obtain the cumulative cost of the candidate path passing through the target neighbor node.

[0053] Obtain the cumulative cost C_accum_u of the path from the root node to Nu from the path tree, and calculate the cumulative cost C_accum_v of the candidate path passing through Nv = C_accum_u + C_total.

[0054] Step S146: For the case where there are different arrival paths for the same target neighbor node, compare the cumulative cost of the candidate paths corresponding to the different arrival paths, retain the arrival path with the smaller cumulative cost of the candidate path, and update the parent node association relationship of the target neighbor node in the path tree as well as the corresponding road segment entry time and road segment exit time.

[0055] If a cumulative cost for a path to Nv already exists in the path tree, compare the existing cumulative cost with the newly calculated cumulative cost of a candidate path. If the cumulative cost of the candidate path is less than the existing cumulative cost, update the parent node pointer of Nv in the path tree to Nu, update the arrival time of Nv to Tv, and update the cumulative cost of Nv to C_accum_v; otherwise, retain the original path information.

[0056] Step S147: Repeat the process of visiting neighboring nodes from the current node and calculating the cumulative cost of candidate paths until the path search termination node is marked as completed. Obtain the path tree branch from the path search start node to the path search termination node. Backtrack from the path search termination node to the path search start node along the path tree branch. Extract the segment unit identifier of all road segment units passed through, the entry time and departure time of each road segment unit, and combine them to generate a complete path segment sequence and the entry time and departure time of each road segment unit in the path segment sequence.

[0057] Repeat steps S143 to S146, using a priority queue to manage the nodes to be visited. Each time, the node with the lowest cumulative path cost is taken from the priority queue as the next current node for expansion. The search is complete when Ne is removed from the priority queue. Starting from Ne, backtrack level by level along the parent node pointers in the path tree to Ns. The sequence of nodes traversed by the backtracking path is the complete path. Extract the segment unit identifiers of the segment units between adjacent nodes in the backtracking path, as well as the entry and exit times of each segment unit, and combine them into the complete path segment sequence P=[E1, E2, ..., Em]. At the same time, output the entry and exit time sequences of each segment unit.

[0058] Step S150: The complete route segment sequence, the entry time and departure time of each segment unit are combined to form the truck route planning and scheduling information, and the truck route planning and scheduling information is sent to the target scheduling terminal to trigger the truck driving route guidance operation.

[0059] The P (entry time sequence) and the departure time sequence are encapsulated into truck route planning and scheduling information, which is then transmitted to the target scheduling terminal in the truck cab via a wireless communication link. Upon receiving the truck route planning and scheduling information, the target scheduling terminal draws the driving route on the display screen, marks the expected entry and departure times for each road segment, and guides the driver to travel along the planned route via voice announcement.

[0060] Step S210: Pre-mark the set of restricted road segment units with truck passage time constraints from the target road network, and extract the start time and end time of the restricted period for each restricted road segment unit.

[0061] Obtain the set of restricted road segment units E_limit from the traffic management system. Each restricted road segment unit in E_limit is associated with a restricted time period, which is defined by the start time T_limit_start and the end time T_limit_end.

[0062] Step S220: In the process of constructing the time-varying hybrid cost road network representation structure, for each restricted road segment unit in the restricted road segment unit set, mark all analysis times within the restricted time period interval as a prohibited state in the hybrid cost time series vector corresponding to the restricted road segment unit.

[0063] When generating the mixed cost time series vector for each restricted road segment unit in step S137, each analysis time ak in the vector is traversed. If T_limit_start≤ak≤T_limit_end, the impedance value field in the mixed cost entry of the road segment at that analysis time is marked as a prohibited passage status identifier.

[0064] Step S230: During the time-varying path search process of the time-varying hybrid cost road network representation structure, when traversing to the restricted road segment unit, the expected entry time of the road segment that the path search reaches the restricted road segment unit is read.

[0065] In step S143, when traversing from Nu to the neighbor node Nv, and Euv belongs to E_limit, the entry time Tv of Nv is read.

[0066] Step S240: Determine whether the expected entry time of the road segment falls within the restricted time period of the restricted road segment unit. When the expected entry time of the road segment falls within the restricted time period, assign a passage prohibition identifier to the restricted road segment unit and temporarily set its corresponding comprehensive path selection cost to an infinite amount of passage cost.

[0067] If T_limit_start≤Tv≤T_limit_end, then Euv is determined to be in a restricted state at time Tv, a passage prohibition identifier is assigned to Euv, and the comprehensive path selection cost C_total in step S144 is set to the infinite passage cost Minf.

[0068] Step S250: When the expected entry time of the road segment does not fall within the restricted time period of the restricted road segment unit, the comprehensive route selection cost is calculated normally according to the road segment impedance value and truck toll coefficient corresponding to the expected entry time of the restricted road segment unit in the time-varying hybrid cost road network representation structure.

[0069] If Tv is less than T_limit_start or greater than T_limit_end, then C_total is calculated according to the normal process of step S144.

[0070] Step S260: During the path search process, a detour avoidance operation is performed on the road segment unit containing the prohibition of passage identifier, so that the time-varying path search process automatically skips the road segment unit in the prohibition of passage state and selects an alternative connecting path.

[0071] In step S146, the path with a total path selection cost of Minf is placed at the end of the priority queue. The dequeue operation of the priority queue automatically selects the alternative connected path with the lower cost to achieve detour avoidance.

[0072] Step S270: After outputting the complete route segment sequence, the complete route segment sequence is reviewed for compliance based on the restricted road segment unit set. It is checked whether the entry time of each road segment unit in the complete route segment sequence overlaps with the restricted time period associated with that road segment unit. The complete route segment sequence that passes the compliance review is used as the route content in the truck route planning and scheduling information, and the prohibited passage status of the restricted road segment unit set is added to the truck route planning and scheduling information.

[0073] Iterate through each road segment unit in P and check whether its entry time falls within its corresponding restricted time period. If all road segment units do not fall within the restricted time period, the compliance review is passed. The P that has passed the compliance review is used as the route content in the truck route planning and scheduling information, and the restricted time period information of each restricted road segment unit in E_limit is added as a prohibited passage status label to the truck route planning and scheduling information.

[0074] Step S310: Starting from the entry time corresponding to the first segment unit of the complete path segment sequence, extract the truck toll fee coefficient and segment impedance value corresponding to each segment unit in the complete path segment sequence in sequence.

[0075] Let the complete path segment sequence P = [E1, E2, ..., Em], and the entry time corresponding to each segment unit be [T1, T2, ..., Tm]. For the i-th segment unit Ei, extract the cost coefficient ci and impedance value zi of Ei at time Ti from the time-varying hybrid cost road network representation structure, using Ti as the time index.

[0076] Step S320: Calculate the total cost contribution of each road segment unit using the freight truck toll coefficient and the geometric length parameter of each road segment unit. Then, sum up the total cost contributions of all road segment units in the complete route sequence to obtain the total route cost.

[0077] Let Li be the geometric length parameter of Ei, and let Fi = ci × Li be the total cost contribution of Ei. The total cost of the path, F_total, is equal to F1 + F2 + ... + Fm.

[0078] Step S330: Calculate the total time consumption contribution of each road segment unit using the road segment impedance value and geometric length parameters of each road segment unit, and then sum up the total time consumption contributions of all road segment units in the complete path road segment sequence to obtain the total path travel time consumption.

[0079] The total time consumption contribution of Ei is Hi = zi × Li. The total time consumption of the path H_total is equal to H1 + H2 + ... + Hm.

[0080] Step S340: Extract the identifiers of all differential toll road segment units with different rates in the target road network from the toll strategy configuration set. Determine whether the complete path road segment sequence contains differential road segment units that belong to the differential toll road segment unit identifiers. If the complete path road segment sequence contains differential road segment units that belong to the differential toll road segment unit identifiers, backtrack the entry time of the differential road segment unit in the complete path road segment sequence. Based on the entry time, obtain the freight truck toll coefficient of the differential road segment unit in the corresponding time-varying hybrid cost road network representation structure as the differential toll reference coefficient.

[0081] Extract the set of road segment unit identifiers E_diff from the toll policy configuration set that contains all road segments with different rates. Iterate through P, and if Ei belongs to E_diff, extract the cost coefficient c_diff_i of Ei at time Ti using Ti as the time index, as the differential cost reference coefficient.

[0082] Step S350: Calculate the cost saving substitution amount for the differentiated road segment using the differentiated cost reference coefficient and the geometric length parameter of the differentiated road segment unit. The cost saving substitution amount for the differentiated road segment represents the change in the cost expenditure of the truck on the differentiated road segment unit relative to the scenario without differentiated tolling. Combine the total route cost expenditure, the total route travel time, and the cost saving substitution amount for the differentiated road segment as independent evaluation dimensions to form multi-dimensional route cost evaluation information, and attach the multi-dimensional route cost evaluation information to the truck route planning and scheduling information.

[0083] Let c_base_j be the base cost coefficient for the differential road segment unit Ej, and let ΔFj be the cost saving substitution amount for the differential road segment ΔFj = (c_diff_j - c_base_j) × Lj. Combine F_total, H_total, and all ΔFj into multi-dimensional path cost evaluation information, and attach it to the truck route planning and scheduling information.

[0084] Step S360: Send the truck route planning and scheduling information carrying multi-dimensional evaluation information of path cost to the target scheduling terminal, triggering the target scheduling terminal to synchronously display the truck driving route guidance operation and multi-dimensional evaluation information of path cost.

[0085] The updated truck route planning and scheduling information in step S350 is sent to the target scheduling terminal. The target scheduling terminal simultaneously displays the total route cost, total route travel time, and cost savings and substitutions for each different route segment on the route display interface.

[0086] Step S410: Obtain historical roadside sensor datasets and historical tolling strategy configuration sets from multiple historical dates within the same planning time interval in the target road network to form a historical traffic scenario sample set. For each set of historical roadside sensor datasets in the historical traffic scenario sample set, use the cross-sectional average traffic flow rate record and cross-sectional traffic flow space occupancy record to perform time-varying impedance assignment processing on all road segment units of the target road network to generate a historical road network impedance distribution state map.

[0087] Extract K historical roadside sensor datasets and historical tolling strategy configuration sets from the historical database, which are within the same planning time interval on historical dates. Perform step S120 on each set of data to generate K sets of historical road network impedance distribution state diagrams.

[0088] Step S420: For each set of historical toll strategy configurations in the historical traffic scenario sample set, a historical time-varying hybrid cost road network representation structure is generated by aligning and projecting the truck toll coefficients in the toll strategy configuration set to the road network impedance distribution state diagram in a time dimension. On each set of historical time-varying hybrid cost road network representation structures, time-varying path search processing is performed for multiple different combinations of virtual truck trip start positions and virtual truck trip end positions to obtain multiple historical complete path segment sequences and the corresponding historical path total cost expenditure and historical path total travel time.

[0089] For each set of historical toll strategy configurations, step S130 is executed to generate K sets of historical time-varying hybrid cost road network representation structures. For Q sets of virtual origin-endpoint combinations, steps S140 and S320 to S330 are executed on each set of historical time-varying hybrid cost road network representation structures to obtain K×Q sets of historical complete path segment sequences and corresponding F_total and H_total.

[0090] Step S430: The historical complete route segment sequences of the same virtual truck trip start position and virtual truck trip end position under different historical dates are clustered according to the similarity of the route segment units to generate typical route clusters representing the corresponding start-end combination. For each typical route cluster, the entry time distribution range and exit time distribution range of each segment are extracted from the corresponding historical complete route segment sequence to generate the trip time band of the typical route cluster. Also, for each typical route cluster, the cost distribution range and trip time distribution range are extracted from the corresponding historical total cost expenditure and historical total trip time to generate the cost distribution range of the typical route cluster.

[0091] For each origin-endpoint combination, hierarchical clustering is performed on its K historical complete path segment sequences based on the Jaccard similarity of the segment units, and the largest cluster is taken as the typical path cluster. For each path within the typical path cluster, the entry and exit times of each segment unit are extracted, and the minimum and maximum values ​​of the entry times of each segment unit are calculated to form the travel time band; the same applies to the exit times. The minimum and maximum values ​​of F_total for the K paths are extracted to form the cost distribution interval, and the minimum and maximum values ​​of H_total are extracted to form the travel time distribution interval.

[0092] Step S440: After performing time-varying path search processing within the current planning period to obtain the current complete path segment sequence, the current complete path segment sequence is matched with the typical path cluster corresponding to the combination of the starting position and ending position of the same truck trip, and it is determined whether the current complete path segment sequence falls within the travel time zone and cost distribution range of the typical path cluster.

[0093] In the current planning cycle, step S140 is executed to obtain the current complete route segment sequence P_curr and the corresponding F_total_curr and H_total_curr. P_curr is matched with the typical route clusters of the corresponding start-end combination. It is checked whether the entry time of each segment unit in P_curr falls within the travel time band, whether F_total_curr falls within the cost expenditure distribution range, and whether H_total_curr falls within the travel time consumption distribution range.

[0094] Step S450: When the deviation direction between the current complete path segment sequence and the typical path cluster indicates that the total path cost of the current complete path segment sequence is too high, extract the historical complete path segment sequence from the typical path cluster as an alternative path suggestion, supplement the truck route planning and scheduling information with the alternative path suggestion, and send the truck route planning and scheduling information with the supplemented alternative path suggestion to the target scheduling terminal.

[0095] If F_total_curr is greater than the upper limit of the cost distribution range, the cost is considered too high. The historical complete route segment sequence P_alt with the smallest F_total is selected from the typical route cluster as an alternative route suggestion, and P_alt is added to the truck route planning and scheduling information and sent to the target scheduling terminal.

[0096] Step S510: Mark the dedicated truck lane network from the target road network. The dedicated truck lane network consists of dedicated road segment units that allow trucks to pass and nodes connecting the dedicated road segment units. The dedicated truck lane network has dedicated lane continuity between the starting position and the ending position of the truck's journey.

[0097] Select a set of road segment units E_spec from the target road network that are marked as dedicated truck lanes. E_spec and the nodes connecting the road segment units in E_spec together constitute the dedicated truck lane network. Check whether E_spec is connected from Ns to Ne.

[0098] Step S520: Extract the dedicated road segment hybrid cost entries corresponding to the dedicated road segment units in the freight dedicated lane network from the time-varying hybrid cost road network representation structure to form a dedicated lane cost substructure.

[0099] Extract the hybrid cost time series vector of each road segment unit in E_spec from the time-varying hybrid cost road network representation structure to form a dedicated channel cost substructure.

[0100] Step S530: Perform coefficient sensitivity classification on the truck toll coefficient corresponding to the dedicated road segment unit in the toll strategy configuration set, and mark the dedicated road segment unit whose deviation from the preset benchmark toll coefficient exceeds the preset deviation threshold as the target toll road segment unit.

[0101] Let the preset baseline cost factor be c_ref. For each road segment unit in E_spec, take the average value of its cost factor for each time period, c_avg, and calculate the deviation range d=|c_avg-c_ref| / c_ref. If d is greater than the preset deviation threshold θ, then mark the road segment unit as the target toll road segment unit E_target.

[0102] Step S540: Before performing time-varying path search processing for the dedicated truck lane network, a cost coefficient copy adjustment is generated for each target toll road segment unit. The cost coefficient copy adjustment is obtained by increasing or decreasing the original truck toll coefficient of the target toll road segment unit.

[0103] For each E_target, generate the adjustment range δ, and the adjusted cost coefficient c_adj=c_avg×(1+δ). When amplification is needed, δ takes a positive value, and when amplification is needed, δ takes a negative value.

[0104] Step S550: Temporarily replace the original truck toll fee coefficient of the target toll road segment unit with the adjusted cost coefficient copy of the target toll road segment unit at the corresponding analysis time, generate the adjusted dedicated lane cost substructure, perform time-varying path search processing on the adjusted dedicated lane cost substructure to obtain the adjusted complete path segment sequence and the corresponding adjusted total path cost expenditure. Based on the complete path segment sequences corresponding to the adjusted dedicated lane cost substructure and the original dedicated lane cost substructure, compare the differences between the two in the total path cost expenditure and the total path travel time to generate a cost coefficient sensitivity influence vector.

[0105] Replace the cost coefficient of E_target in the dedicated channel cost substructure with c_adj to obtain the adjusted dedicated channel cost substructure. Execute step S140 to obtain the adjusted complete path segment sequence P_adj, F_total_adj, and H_total_adj. Calculate ΔF_total = F_total_adj - F_total_orig and ΔH_total = H_total_adj - H_total_orig, and combine them into a cost coefficient sensitivity influence vector.

[0106] Step S560: The cost factor sensitivity influence vector is compared with the complete path segment sequence obtained by performing time-varying path search processing on the time-varying hybrid cost road network representation structure. The list of road segment units sensitive to cost factor changes in the complete path segment sequence is extracted. The list of road segment units sensitive to cost factor changes and the cost factor sensitivity influence vector are combined into the cost sensitivity analysis result. The cost sensitivity analysis result is then added to the truck route planning and scheduling information.

[0107] P_adj and P_orig are compared one by one according to road segment units. The road segment units corresponding to the positions where the road segment unit identifiers are different are extracted and used as a list of road segment units sensitive to changes in cost coefficients. This list and the cost coefficient sensitivity influence vector are combined to form the cost sensitivity analysis result and added to the truck route planning and scheduling information.

[0108] Step S570: Send the truck route planning and scheduling information carrying the cost sensitivity analysis results to the target scheduling terminal, triggering the target scheduling terminal to simultaneously display a list of road segment units sensitive to changes in cost coefficients when displaying truck driving route guidance operations.

[0109] The updated truck route planning and scheduling information from step S560 is sent to the target scheduling terminal. While displaying the driving route, the target scheduling terminal highlights a list of road segment units that are sensitive to changes in cost coefficients.

[0110] Step S610: Incremental roadside sensor data and incremental toll policy update records are collected from the roadside sensor dataset and toll policy configuration set of the target road network at a preset real-time refresh cycle. Time-varying impedance assignment processing is performed on the newly added cross-section average traffic flow rate record and newly added cross-section traffic flow space occupancy record in the incremental roadside sensor data to generate incremental road network impedance distribution update segment.

[0111] The newly added data records whose t is greater than the last update time are collected at a refresh period of T_refresh. Step S120 is executed on the incremental data to obtain the incremental road network impedance distribution update segment.

[0112] Step S620: Perform time dimension alignment projection processing on the newly added truck toll fee coefficient and the switching time sequence of the newly added fee coefficient in the incremental toll strategy update record to generate an incremental hybrid cost update fragment. Use the incremental road network impedance distribution update fragment and the incremental hybrid cost update fragment to locally replace the affected road segment units and time segment range in the time-varying hybrid cost road network representation structure to obtain the rolling update time-varying hybrid cost road network representation structure.

[0113] Step S130 is executed on the incremental tolling strategy update record to obtain the incremental hybrid cost update segment. In the time-varying hybrid cost road network representation structure, the road segment units and time segment ranges affected by the incremental data are located, and the original data at the corresponding positions is replaced with incremental data to obtain the rolling updated time-varying hybrid cost road network representation structure.

[0114] Step S630: Map the actual position of the truck at the current moment to the rolling updated time-varying hybrid cost road network representation structure as the starting node of replanning, and use the end position of the truck's journey as the ending node of replanning.

[0115] The truck's current location is obtained from its onboard positioning device and mapped to the nearest node in the target road network as the starting node for replanning. The truck's destination location remains unchanged.

[0116] Step S640: Perform time-varying path search processing on the rolling updated time-varying hybrid cost road network representation structure to generate a replanned complete path segment sequence and the replanning entry time and replanning exit time of each segment unit in the replanned complete path segment sequence.

[0117] Using the replanning start node and replanning end node as input, step S140 is performed on the rolling updated time-varying hybrid cost road network representation structure to obtain the replanning complete path segment sequence.

[0118] Step S650: Calculate the degree of path deviation between the replanned complete path segment sequence and the original complete path segment sequence, determine the degree of consistency between the replanned complete path segment sequence and the original complete path segment sequence in terms of segment unit composition, and when the degree of consistency indicates that the path has deviated, extract the newly added and removed segment units in the replanned complete path segment sequence to generate a list of path change differences.

[0119] Calculate the Jaccard similarity of road segment units between the replanned complete route segment sequence and the original complete route segment sequence. If the similarity is less than a preset threshold, extract road segment units that are present in the replanned complete route segment sequence but not in the original complete route segment sequence as new road segment units; otherwise, remove road segment units to form a list of route change differences.

[0120] Step S660: The replanned complete route segment sequence, replanned entry time, replanned departure time, and route change difference list are combined to form updated route planning and scheduling information, and the updated route planning and scheduling information is sent to the target scheduling terminal to trigger dynamic adjustment guidance of the truck's driving route.

[0121] The replanning results and the list of route change differences are packaged into updated route planning and scheduling information and sent to the target scheduling terminal. The target scheduling terminal uses different colors to mark the newly added and removed road segment units according to the list of route change differences, guiding drivers to adjust their driving routes.

[0122] For example, the method may also include: step S710, reading the road segment geometric curvature parameters and road segment longitudinal slope parameters of each road segment facility in the target road network to form a set of road segment geometric characteristic parameters; and reading the truck's rated gross vehicle weight parameters and truck engine power parameters to form a set of truck power characteristic parameters.

[0123] The geometric curvature parameter *r* and longitudinal slope parameter *s* of each road segment unit are read from the geographic information database of the target road network to form a set of road segment geometric characteristic parameters. The rated gross vehicle weight parameter *m* and engine power parameter *p_w* of the trucks are read from the truck vehicle information database to form a set of truck dynamic characteristic parameters.

[0124] Step S720: Construct the road segment traffic resistance influence coefficient distribution using road segment geometric curvature parameters and road segment longitudinal slope parameters. The road segment traffic resistance influence coefficient distribution maps the geometric characteristics of the road segment unit to the additional energy consumption influence of trucks traveling in the road segment unit.

[0125] A function g(r, s) representing the influence coefficient of road segment traffic resistance is constructed, where g increases with increasing curvature and slope. One implementation is g = 1 + ε × r + ζ × s, where ε is the curvature resistance coefficient and ζ is the slope resistance coefficient. The resistance influence coefficient g is calculated for each road segment unit, forming the distribution of road segment traffic resistance influence coefficients.

[0126] Step S730: Construct a truck power output characteristic curve using the truck's rated gross vehicle weight parameters and truck engine power parameters. The truck power output characteristic curve maps the truck's power characteristics to the truck's energy consumption per unit mileage at different driving speeds.

[0127] Construct the power output characteristic function of the truck e(v, m, p_w). e follows a U-shaped curve that first decreases and then increases as v increases. The larger m is, the higher e generally becomes, and the larger p_w is, the lower e generally becomes.

[0128] Step S740: The distribution of road segment traffic resistance influence coefficients is superimposed with the truck power output characteristic curve to generate an adaptive energy consumption mapping relationship for specific truck power characteristics and specific road segment geometric characteristics. During the time-varying path search process on the time-varying hybrid cost road network representation structure, for each traversed road segment unit, while reading the road segment impedance value and truck traffic cost coefficient, the unit mileage energy consumption estimate corresponding to the road segment geometric curvature parameter and road segment longitudinal slope parameter of the road segment unit is obtained from the adaptive energy consumption mapping relationship.

[0129] The adaptive energy consumption mapping relationship is defined as u(v, r, s, m, p_w) = e(v, m, p_w) × g(r, s). When executing step S144, for road segment unit Ei, the estimated driving speed v_est at the current time is used as input, and the estimated energy consumption per unit mileage u_i corresponding to Ei is obtained from u.

[0130] Step S750: Calculate the total energy consumption of the road segment unit using the estimated energy consumption per unit mileage and the geometric length parameter of the road segment unit. Convert the total energy consumption into an energy cost component using a preset energy cost coefficient. Convert the road segment impedance value into a time cost component using a preset time value coefficient. Use the truck toll coefficient as a monetary cost component. Input the energy cost component, time cost component, and monetary cost component into the integrated path selection cost fusion function. The integrated path selection cost fusion function performs a weighted summation of the energy cost component, time cost component, and monetary cost component to generate an integrated path selection cost that includes energy consumption factors and is expressed in a unified monetary dimension.

[0131] The total energy consumption of Ei is U_i = u_i × L_i. The energy cost component is C_energy = U_i × η, where η is a preset energy cost coefficient, converting energy consumption into monetary units. The comprehensive path selection cost is C_total = w1 × C_time + w2 × C_fee + w3 × C_energy, where w1, w2, and w3 are preset weighting coefficients, and C_total is in monetary units.

[0132] Step S760: After obtaining the complete route segment sequence, extract the total energy consumption corresponding to each segment unit in the complete route segment sequence and sum them up one by one to obtain the total energy consumption of the route. Then, associate and combine the total energy consumption of the route with the complete route segment sequence in the truck route planning and scheduling information to form truck route planning and scheduling information carrying energy consumption information. Finally, send the truck route planning and scheduling information carrying energy consumption information to the target scheduling terminal.

[0133] The total energy consumption along the route, U_total, equals U_1 + U_2 + ... + U_m. U_total is associated with and combined with P to form truck route planning and scheduling information carrying energy consumption data, which is then sent to the target scheduling terminal.

[0134] Step S810: Obtain historical roadside sensor datasets and historical toll strategy configuration sets for the target road network in multiple historical planning periods. Perform time-varying impedance assignment processing and time dimension alignment projection processing on each set of historical roadside sensor datasets and historical toll strategy configuration sets to generate corresponding historical time-varying hybrid cost road network representation structure samples. From each historical time-varying hybrid cost road network representation structure sample, extract the road segment impedance values ​​and truck toll coefficients of all road segment units at all analysis times to form a full cost snapshot of the road network. Mark the actual truck route selection records that occurred in the historical planning period on the full cost snapshot of the road network. The truck route selection records include the actual truck trip start position, the actual truck trip end position, and the complete route segment sequence actually traveled.

[0135] Historical data from multiple planning periods are extracted from the historical database, and steps S120 and S130 are executed to generate historical time-varying hybrid cost road network representation structure samples. All (R, t, z, c) quadruplets are extracted from each historical time-varying hybrid cost road network representation structure sample to form a full cost snapshot of the road network. Historical truck trajectory data are associated and labeled with the full cost snapshot of the road network; each labeled record contains the true start-point location, the true end-point location, and the true path segment sequence.

[0136] Step S820: Using the full cost snapshot of the road network as input and the complete sequence of road segments actually traveled as output target, a pre-defined deep path preference learning network is trained, so that the deep path preference learning network learns the mapping relationship between the full cost snapshot of the road network and the route selection behavior of trucks. The deep path preference learning network includes a graph attention encoding module and a path sequence decoding module. The graph attention encoding module is used to perform multi-round attention message passing on the road network graph structure with road segment units as nodes and road segment connection relationships as edges to generate the implicit preference representation vector of each road segment unit. The path sequence decoding module is used to gradually generate the path segment sequence based on the implicit preference representation vectors of the start node and the end node.

[0137] The deep path preference learning network consists of two modules. The graph attention encoding module takes the road network structure as input. Each road segment unit node uses a vector concatenated from its segment impedance values ​​and truck toll coefficients across all analysis times as its initial feature. Through a multi-round graph attention mechanism, messages are passed and aggregated between adjacent nodes to generate the latent preference representation vector for each road segment unit. The path sequence decoding module takes the latent preference representation vectors of the start and end nodes as input and uses a recurrent neural network architecture to progressively generate the path segment sequence. At each step, based on the latent preference representation vectors of the currently generated partial paths and candidate connected road segment units, the probability distribution for selecting each road segment in the next step is calculated. The real path segment sequence is used as the supervision target, and the cross-entropy loss function is employed for training.

[0138] In step S830, during the training process, when the path sequence decoding module generates the path segment sequence step by step, the path segment sequence generated at each step in the step-by-step generation process is used as the historical trajectory context on which the next generation is based. The historical trajectory context is encoded through a self-attention mechanism to obtain the trajectory context vector. The trajectory context vector is then aggregated with the implicit preference representation vector of the next candidate road segment unit to determine the probability distribution of selecting each road segment in the next step.

[0139] The path sequence decoding module is implemented as follows: At each generation step, the generated partial path segment sequence is used as the historical trajectory context. The attention weight distribution of each segment in the historical trajectory context is calculated using a self-attention mechanism, and the weighted sum is obtained to obtain the trajectory context vector. The trajectory context vector is then aggregated with the implicit preference representation vectors of all outgoing connected segment units of the current node. The aggregation result is mapped through a fully connected layer to the probability distribution of selecting each segment in the next step. The segment unit with the highest probability is selected as the result for the next step.

[0140] Step S840: Construct a joint sensitivity branch network for road segment impedance and rate. The joint sensitivity branch network for road segment impedance and rate receives the implicit preference representation vector output by the graph attention coding module, and outputs the impedance preference adjustment amount and rate preference adjustment amount of each road segment unit at a given analysis time through parallel road segment impedance sensitivity fully connected layer and rate sensitivity fully connected layer.

[0141] Two independent fully connected layer branches are connected in parallel at the output of the graph attention encoding module. One branch is a segment impedance sensitivity fully connected layer, which receives the latent preference representation vector and outputs the impedance preference adjustment amount; the other branch is a rate sensitivity fully connected layer, which receives the latent preference representation vector and outputs the rate preference adjustment amount. The output dimension of the two fully connected layers is consistent with the number of segment units, and each component represents the adjustment amount for one segment unit.

[0142] In step S850, the impedance preference adjustment amount and rate preference adjustment amount corresponding to the time cost component and the monetary cost component, respectively, are injected as additional bias terms into the path sequence decoding module of the deep path preference learning network to correct the response sensitivity of the path sequence decoding module to the path selection cost component that has been converted to a unified dimension during the generation of path segment sequences. After training is completed, the full cost snapshot of the road network corresponding to the time-varying hybrid cost road network representation structure generated in the current planning period is input into the trained deep path preference learning network, and the current latent preference representation vector of each road segment unit is generated through the graph attention encoding module.

[0143] When the path sequence decoding module calculates the probability distribution for the next step, the impedance preference adjustment is superimposed on the scoring item corresponding to the time cost component of the candidate road segment unit, and the rate preference adjustment is superimposed on the scoring item corresponding to the monetary cost component to correct the score. After training, a snapshot of the full road network cost for the current planning period is input into the deep path preference learning network, and the graph attention encoding module generates the current latent preference representation vector.

[0144] Step S860: The current implicit preference representation vector is processed through the joint sensitivity branch network of road segment impedance and rate, and the current impedance preference adjustment and current rate preference adjustment of each road segment unit at the current analysis time are output. The current impedance preference adjustment and current rate preference adjustment are injected into the path sequence decoding module. Combined with the starting node corresponding to the given truck trip start position and the ending node corresponding to the truck trip end position, the path sequence decoding module generates each step of the path segment sequence step by step. At each step, the trajectory context vector of the generated current part of the path segment sequence and the current implicit preference representation vector of the candidate road segment unit are fused using the self-attention mechanism, and the corresponding current impedance preference adjustment and current rate preference adjustment are superimposed to generate the path segment selection decision for each step. Finally, the preference learning path segment sequence and the preference entry time and preference exit time of each road segment unit are obtained.

[0145] The current implicit preference representation vector is input into the joint impedance and toll rate sensitivity branch network to obtain the current impedance preference adjustment and the current toll rate preference adjustment. Using the start and end nodes as inputs, the path sequence decoding module generates a path step by step. At each step, the trajectory context vector and the current implicit preference representation vector of the candidate road segment unit are fused. The current impedance preference adjustment and the current toll rate preference adjustment are then superimposed to calculate the probability distribution. The maximum value is taken as the selection for each step until the end node is reached. The preferred entry time and preferred exit time are calculated based on the estimated travel time of each road segment unit in the generated path.

[0146] Step S870: Compare the path segment sequence of the preference learning path with the complete path segment sequence obtained through time-varying path search processing to generate path preference deviation analysis results. The path preference deviation analysis results are used to describe the differences in segment composition and travel time between the path generated based on historical truck behavior preferences and the path generated based on hybrid cost minimum.

[0147] The Jaccard similarity of road segment units between the preferred learning path segment sequence and the complete path segment sequence is calculated. Different road segment units are extracted from the two, and the number and proportion of different road segment units are counted. The travel time difference between the two paths is compared and combined into the path preference deviation analysis results.

[0148] Step S880: The path preference deviation analysis results and the preference learning path segment sequence are appended to the truck route planning and scheduling information, and the appended truck route planning and scheduling information is sent to the target scheduling terminal.

[0149] The path preference deviation analysis results and the path segment sequence of preference learning are added as supplementary information to the truck route planning and scheduling information and sent to the target scheduling terminal. The target scheduling terminal can switch between displaying the path based on the minimum cost and the path based on historical preferences on the interface.

[0150] Step S910: Deploy multiple virtual toll rate adjustment points in the target road network. Each virtual toll rate adjustment point corresponds to a road segment unit or a set of multiple consecutive road segment units. Configure an adjustable virtual toll rate adjustment variable for each virtual toll rate adjustment point. The value range of the virtual toll rate adjustment variable is constrained by the upper and lower fluctuation range of the original truck toll rate coefficient of the road segment unit corresponding to the virtual toll rate adjustment point.

[0151] Several road segment units are pre-selected in the target road network as virtual rate adjustment points. Each virtual rate adjustment point j corresponds to a virtual rate adjustment variable δj, and the value range of δj is [-δmax, δmax]. The actual adjusted cost coefficient is c_actual=c_orig×(1+δj).

[0152] Step S920: Construct a rate adjustment decision agent based on a policy network. The rate adjustment decision agent receives the time-varying hybrid cost road network representation structure within the current planning period as the environmental state input, and extracts the traffic impedance characteristics and rate distribution characteristics of each road segment unit in the time-varying hybrid cost road network representation structure in the spatiotemporal dimension through multi-layer graph convolution operation.

[0153] The rate adjustment decision agent based on the policy network takes a time-varying hybrid cost road network representation as input. First, a feature vector is constructed for each road segment unit, containing the road segment impedance value and rate coefficient at each analysis time. Features are extracted through multi-layer graph convolution operations. The first layer of graph convolution takes the road segment unit adjacency matrix and the initial feature matrix as input, and outputs the first hidden layer feature matrix. The second layer of graph convolution takes the first hidden layer feature matrix as input and outputs the second hidden layer feature matrix. The multi-layer graph convolution operation aggregates the neighborhood traffic impedance features and rate distribution features of each road segment unit.

[0154] Step S930: The extracted traffic impedance features and rate distribution features are input into the strategy output layer of the rate adjustment decision agent. The strategy output layer outputs a rate adjustment action probability distribution for each virtual rate adjustment point. The rate adjustment action probability distribution includes the execution probability of rate increase action, rate decrease action and rate maintenance action respectively.

[0155] The feature vectors of the road segment units corresponding to each virtual rate adjustment point in the second hidden layer feature matrix are input into the policy output layer. The policy output layer is a fully connected layer with three output nodes plus a flexible maximum transfer function. The three output nodes correspond to the execution probabilities of the three actions: adjustment, reduction and maintenance.

[0156] Step S940: Sample a specific rate adjustment action for each virtual rate adjustment point according to the probability distribution of the adjustment action, and adjust the virtual rate adjustment variable corresponding to the virtual rate adjustment point according to the sampled rate adjustment action to generate an updated virtual toll strategy configuration set; in the updated virtual toll strategy configuration set, the freight truck toll coefficient of the road segment unit affected by the rate adjustment action is shifted from the original value.

[0157] The execution action for each virtual toll rate adjustment point is obtained by sampling probabilistically from the probability distribution of adjustment actions. If the sampling result is an increase, δj is increased by a preset step size; if it is a decrease, δj is decreased by a preset step size; if it is to remain unchanged, δj remains unchanged. The cost coefficient of the adjusted road segment unit is updated to generate a virtual toll strategy configuration set.

[0158] Step S950: Replace the original toll strategy configuration set with the updated virtual toll strategy configuration set, and repeatedly execute the time dimension alignment projection processing and time-varying path search processing to generate a virtual path segment sequence under the virtual toll strategy, as well as the corresponding total virtual path cost and total virtual path travel time.

[0159] Replace the original toll policy configuration set with the virtual toll policy configuration set, and execute steps S130 and S140 to obtain the virtual path segment sequence and F_total_virtual and H_total_virtual.

[0160] Step S960: Statistically analyze the traffic load increment changes of each segment unit in the virtual route segment sequence, and generate a road network traffic load redistribution feature map based on the spatial distribution pattern of traffic load increment changes of multiple consecutive adjacent segment units. The road network traffic load redistribution feature map is used to characterize the direction and magnitude of the impact of truck route shifts caused by virtual rate adjustments on the traffic pressure of each segment unit of the road network.

[0161] The traffic volume difference between the virtual path and the original path for each road segment unit is calculated as the traffic load increment change. The traffic load increment change of all road segment units is plotted as a traffic load redistribution characteristic map of the road network according to the spatial topology.

[0162] Step S970: The deviation between the road network traffic load redistribution characteristic map and the preset road network load balance target state is calculated. At the same time, the direction and magnitude of the change in the total cost of the virtual path and the total travel time of the virtual path relative to the corresponding indicators of the path under the original strategy are evaluated. Based on the comprehensive deviation calculation results and the evaluation of the changes in cost and time indicators, a rate adjustment effect evaluation feedback signal is generated. The rate adjustment effect evaluation feedback signal outputs a positive feedback value when the virtual rate adjustment makes the road network traffic load distribution approach the preset road network load balance target state and the change in cost or time indicators meets the preset control expectation; otherwise, it outputs a negative feedback value.

[0163] Calculate the mean square error (MSE_BAL) between the road network traffic load redistribution characteristic map and the road network load balance target state. Calculate ΔF = F_total_virtual - F_total_orig and ΔH = H_total_virtual - H_total_orig. If MSE_BAL is less than the preset balance threshold and ΔF and ΔH do not exceed their respective preset tolerance limits, output a positive feedback value +1; otherwise, output a negative feedback value -1.

[0164] Step S980: Update the strategy output layer parameters of the rate adjustment decision agent using the feedback signal of rate adjustment effect evaluation. Optimize the strategy output layer by maximizing the probability of adjustment actions corresponding to positive feedback values, so that the rate adjustment decision agent can gradually learn and generate rate adjustment strategies that can guide the distribution of road network traffic load towards equilibrium.

[0165] Using the feedback signal of the rate adjustment effect evaluation as the reward signal, the policy gradient algorithm is used to update the weight parameters of the fully connected layer of the policy output layer. The update direction is to increase the probability of sampling action when positive feedback is generated.

[0166] Step S990: After the rate adjustment decision agent completes training, the time-varying hybrid cost road network representation structure generated in real time within the current planning period is input into the trained rate adjustment decision agent, and a set of real-time rate adjustment action suggestions for the current planning period is output. The set of real-time rate adjustment action suggestions includes the suggested adjustment action and adjustment range corresponding to each virtual rate adjustment point.

[0167] The currently generated time-varying hybrid cost road network representation structure is input into the trained rate adjustment decision agent. The policy output layer outputs the probability distribution of adjustment actions for each virtual rate adjustment point. The action with the highest probability is taken as the suggested adjustment action. The adjustment range is determined based on the cumulative adjustment amount, and the results are combined into a set of real-time rate adjustment action suggestions.

[0168] Step S9100: The set of real-time rate adjustment action suggestions is injected into the truck route planning and scheduling information as additional strategy suggestion information, and the injected truck route planning and scheduling information is sent to the target scheduling terminal, triggering the target scheduling terminal to display the predicted route transfer trend and road network load change prediction under the condition of implementing the set of real-time rate adjustment action suggestions.

[0169] The set of real-time rate adjustment action suggestions is attached to the truck route planning and scheduling information and sent to the target scheduling terminal. The target scheduling terminal displays on the interface the possible routes that trucks may transfer after the above rate adjustment suggestions are implemented and the load change prediction map of each road segment unit of the road network.

[0170] The deep path preference learning network involved in step S820 has the following specific architecture: This deep path preference learning network adopts an encoder-decoder architecture. The encoder is a graph attention network, containing three stacked graph attention layers. Each graph attention layer is configured with four attention heads, and the hidden layer dimension of each attention head is 64. Residual connections and layer normalization are used between layers. The decoder is a sequence generation structure based on a long short-term memory network, containing two long short-term memory network layers. Each hidden layer has a hidden layer dimension of 128. The decoder ends with a fully connected output layer, the output dimension of which is equal to the total number of road segment units in the target road network. A flexible maximum transfer function is then used to generate the road segment selection probability distribution.

[0171] The technical implementation of each module in the graph attention encoding module is as follows: Each graph attention layer performs multi-head self-attention computation on the input road segment unit feature matrix. For each attention head, the feature vector hi of road segment unit node i is linearly transformed into query vector qi, key vector ki, and value vector vi through learnable weight matrices WQ, WK, and WV, respectively. The attention coefficient eij between node i and its neighboring node j is calculated as the inner product of qi and kj divided by the square root of the key vector dimension, and normalized by the flexible maximum transfer function to obtain the attention weight αij. The value vectors of all neighboring nodes are weighted and summed according to the attention weights to obtain the aggregate vector of node i under this attention head. The aggregate vectors of multiple attention heads are concatenated along the feature dimension, transformed by a feedforward neural network, and then residually connected and layer normalized with the input to output the hidden layer features of the graph attention layer. The three graph attention layers are stacked sequentially, with the output of the previous layer serving as the input of the next layer, ultimately outputting the latent preference representation vector of each road segment unit.

[0172] The technical implementation of each module in the path sequence decoding module is as follows: The initial state of the decoder is concatenated by linear transformation of the latent preference representation vectors of the start and end nodes, and mapped to the initial cell state and initial hidden state of the Long Short-Term Memory (LSTM) network through a fully connected layer. During each generation step, the latent preference representation vector of the last segment unit in the currently generated path segment sequence is used as the input for the current step. After processing through two layers of LSM networks, the hidden output vector of the current step is obtained. The historical trajectory context is implemented through a multi-head self-attention module. This module takes the latent preference representation vectors of all segment units in the generated path segment sequence as input, calculates the attention weight of each segment unit relative to the current generation step, and sums them to obtain the trajectory context vector. The hidden output vector of the current step is concatenated with the trajectory context vector along the feature dimension, mapped to the subsequent candidate segment unit space through a fully connected layer, and then aggregated with the latent preference representation vectors of all candidate segment units through a dot product attention function. The probability distribution of the next segment selection is output through a flexible maximum transfer function.

[0173] The technical implementation of each module in the joint impedance and rate sensitivity branch network is as follows: This branch network receives the latent preference representation vector output by the graph attention coding module and divides it into two parallel fully connected branches. The impedance sensitivity branch contains two fully connected layers. The first fully connected layer has the input dimension of the latent preference representation vector and an output dimension of 128, with a linear rectified function as the activation function. The second fully connected layer has an output dimension of 1, no activation function, and outputs the impedance preference adjustment amount for each road segment unit. The rate sensitivity branch adopts the same structure and outputs the rate preference adjustment amount for each road segment unit.

[0174] The training data and process of the deep path preference learning network are as follows: The training data comes from historical roadside sensor datasets and historical toll policy configuration sets stored in the historical traffic database for multiple historical planning periods. After processing in steps S120 and S130, historical time-varying hybrid cost road network representation structure samples are generated, and a preset number of full cost snapshot samples of the road network are extracted. Each sample is labeled with the actual truck route selection record within the historical planning period as the output target. The input is the full cost snapshot of the road network, and the output is the complete sequence of road segments actually traveled. The training uses the cross-entropy loss function to calculate the cross-entropy error between the predicted probability distribution of road segment selection in each generation step and the actual road segment selection. The sum of the errors of all generation steps is used as the total loss. The optimizer uses the adaptive moment estimation optimizer, with the initial learning rate, batch size, and number of training rounds set to preset values. During training, the average absolute path similarity on the validation set is used as the evaluation index for early model termination. Training stops when this index does not improve in a preset number of consecutive rounds.

[0175] The deep path preference learning network model is applied as follows: During the inference phase, the time-varying hybrid cost road network representation structure generated in the current planning cycle is converted into a full cost snapshot format. The feature vectors of each road segment unit in the full cost snapshot are standardized using the feature mean vector and feature standard deviation vector saved during the training phase before input. The standardized feature vectors are then input into the graph attention encoding module to generate the current latent preference representation vector. Given the starting node corresponding to the truck's journey start point and the ending node corresponding to the truck's journey end point, the path sequence decoding module generates a path segment sequence step by step, starting from the starting node. After generating each road segment unit, it is added to the already generated partial path segment sequence as the historical trajectory context for the next generation. The generation process terminates when the ending node is reached or the number of generation steps exceeds the preset maximum number of steps. The entry and exit times of each road segment unit in the generated path segment sequence are calculated based on the predicted path length, estimated travel speed, and timestamp.

[0176] The rate adjustment decision agent based on the policy network involved in step S920 has the following specific architecture: This rate adjustment decision agent adopts a structure combining a graph convolutional network and a multi-layer fully connected network. The graph convolutional network consists of two graph convolutional layers, each with an output feature dimension of 64 and an activation function of linear rectified function. Following the graph convolutional layers is a fully connected hidden layer with an output dimension of 128 and an activation function of linear rectified function. The policy output layer is a fully connected layer, with an output dimension equal to the number of virtual rate adjustment points multiplied by the product of the dimensions of the three discrete actions corresponding to each adjustment point, followed by a flexible maximum transfer function to convert it into the probability distribution of each action.

[0177] The technical implementation of each module of the rate adjustment decision agent based on the policy network is as follows: The first layer of the graph convolutional network takes the road network adjacency matrix and the road segment unit input feature matrix as input. Each row of the road segment unit input feature matrix is ​​a feature vector of a road segment unit. The feature vector is formed by concatenating the time series of the road segment impedance value and the time series of the rate coefficient of the road segment unit along the time dimension and then standardizing it. The graph convolution calculation is implemented as follows: For each road segment unit node, its own features are weighted and aggregated with the features of the first-order neighbor nodes indexed by the adjacency matrix. The aggregation weight is determined by the Laplace normalization form of the adjacency matrix. The aggregation result is output as the first hidden layer after linear transformation of the learnable weight matrix and activation by the linear rectified function. The second layer of graph convolution takes the output of the first hidden layer as input and repeats the above aggregation and transformation process to output the second hidden layer features. The feature vectors of the road segments where each virtual rate adjustment point is located are extracted from the second hidden layer features, concatenated along the feature dimension, and then transformed by a fully connected hidden layer. The strategy output layer groups the output of the fully connected hidden layer according to the virtual rate adjustment point, and each group calculates the probability distribution of the three actions of adjustment, down adjustment and maintenance through an independent flexible maximum transfer function.

[0178] The training data and process of the rate adjustment decision agent based on the policy network are as follows: Training adopts the proximal policy optimization method in the policy gradient algorithm for online interactive training. In each training round, the time-varying hybrid cost road network representation structure of the current planning period is used as the environmental state input. The rate adjustment decision agent outputs the probability distribution of adjustment actions and samples them for execution. After generating a virtual toll policy configuration set, the rate adjustment effect evaluation feedback signal is obtained through simulation in steps S940 to S970. To improve training efficiency, a parallel simulation method is adopted, and multiple simulation processes are started simultaneously. Each process is executed in an independent virtual environment copy. Each process transmits the state, action, and feedback signal samples back to the training process for centralized gradient update. The update magnitude of the policy network parameters is limited by the pruning objective function, and the ratio of the new and old policies is constrained within a preset range to balance exploration and utilization.

[0179] The application of the policy network-based rate adjustment decision agent model is as follows: After training, the policy network parameters saved during training are loaded into the online inference process. In the real-time application phase, the time-varying hybrid cost road network representation structure generated in real time during the current planning cycle is standardized and input into the rate adjustment decision agent. The policy output layer outputs the probability distribution of adjustment actions corresponding to all virtual rate adjustment points at once. For each virtual rate adjustment point, the action with the highest probability is selected as the recommended action, and the action probability value is used as the confidence score of the action. The recommended actions are output in list form, and each record in the list includes the virtual rate adjustment point identifier, the recommended action type, and the confidence score. This list is the real-time rate adjustment action suggestion set, which is then encapsulated by downstream modules into the truck route planning and scheduling information.

[0180] In one exemplary embodiment, a truck route planning system integrating dynamic traffic flow and differentiated toll policies is provided. This system can be a terminal, server, etc., and its internal structure diagram can be as follows: Figure 2 As shown, this truck routing system integrating dynamic traffic flow and differentiated pricing policies includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, near-field communication, or other technologies. When the computer program is executed by the processor, it implements a truck routing method integrating dynamic traffic flow and differentiated pricing policies. The display unit is used to generate a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, or a button, trackball, or touchpad set on the casing of a truck routing system that integrates dynamic traffic flow and differentiated pricing policies. It can also be an external keyboard, touchpad, or mouse, etc.

[0181] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A truck route planning method integrating dynamic traffic flow and differentiated toll policies, characterized in that, The method includes: The system receives roadside sensor datasets and toll policy configuration sets for the target road network within the planned time interval. The roadside sensor datasets include average traffic flow rate records and traffic flow space occupancy records for each road segment unit at multiple collection times. The toll policy configuration sets include truck toll coefficients and toll coefficient switching time sequences for each road segment unit within multiple time intervals. The average traffic flow rate record and the traffic flow space occupancy rate record of the cross section are used to perform time-varying impedance assignment processing on all road segment units of the target road network to generate a road network impedance distribution state map that is dynamically updated with time segments. The road segment impedance value in the road network impedance distribution state map is determined by fusing the traffic flow state mapping relationship of the average traffic flow rate record and the traffic flow space occupancy rate record of the cross section. The truck toll coefficients in the toll strategy configuration set are projected onto the road network impedance distribution state map in a time dimension alignment process, so that each road segment unit carries the road segment impedance value and the corresponding truck toll coefficient at each analysis step in the time series, forming a time-varying hybrid cost road network representation structure. Based on the given starting and ending points of the truck trip, a time-varying path search process is performed on the time-varying hybrid cost road network representation structure. The time-varying path search process reads the road segment impedance value and truck passage cost coefficient corresponding to the entry time of the road segment when traversing adjacent road segment units, and combines them to calculate a comprehensive path selection cost. The complete path segment sequence and the entry and exit times of each road segment unit in the path segment sequence are output. The complete route segment sequence, the entry time and the departure time of each segment unit are combined to form truck route planning and scheduling information, and the truck route planning and scheduling information is sent to the target scheduling terminal to trigger truck driving route guidance operation.

2. The truck route planning method integrating dynamic traffic flow and differentiated toll policies according to claim 1, characterized in that, The process of assigning time-varying impedance values ​​to all road segment units of the target road network using the cross-sectional average traffic flow rate record and the cross-sectional traffic flow space occupancy record, and generating a road network impedance distribution state map that is dynamically updated with time segments, includes: Extract the average speed records of cross-sectional traffic flow of the target road segment unit in the roadside sensor dataset at continuous acquisition time, and generate an average speed time series. Extract the cross-sectional traffic flow space occupancy rate records of the target road segment unit in the roadside sensor dataset at the continuous acquisition time, and generate a space occupancy rate time series. The average rate time series and the space occupancy time series at the same collection time are input into the traffic flow state mapping relationship to calculate the segment impedance value of the target segment unit at that collection time. The traffic flow state mapping relationship outputs an increased segment impedance value when the occupancy sampling value increases and the rate sampling value decreases, and outputs a decreased segment impedance value when the occupancy sampling value decreases and the rate sampling value increases. Traverse all road segment units included in the target road network, calculate the road segment impedance value of each road segment unit at each collection time according to the traffic flow state mapping relationship, and form a road segment unit impedance time series set. Each element in the road segment unit impedance time series set uniquely corresponds to the road segment unit's road segment impedance value at a collection time. The road segment impedance values ​​in the time series set of road segment unit impedance are organized according to the spatial topological connection relationship of the road segment units. Based on the node adjacency list structure of the target road network, a road network impedance distribution state map with road segment units as the basic expression unit is constructed. The road network impedance distribution state map generates an impedance snapshot layer that is isomorphic to the topological structure of the target road network at each acquisition time. Impedance transition filling is performed on the time gap between adjacent acquisition times. The road segment impedance value between adjacent acquisition times is generated by using the road segment impedance value of the previous acquisition time and the road segment impedance value of the next acquisition time through a time linear transition method. This makes the road network impedance distribution state map form an impedance change flow that is continuously and dynamically updated with time segments within the planned time interval. The impedance values ​​of each road segment unit in the impedance change flow at the transition time in different time segments are associated and stored with the corresponding road segment unit identifier and timestamp mark to obtain a time-refined road network impedance representation carrying the spatial and temporal attribute information of the road segment unit. The transition time impedance values ​​of all road segment units belonging to the same time segment in the time-refined road network impedance representation are integrated into a time segment impedance distribution surface. The time segment impedance distribution surfaces of multiple consecutive time segments together form a road network impedance distribution state map that is dynamically updated with time segments.

3. The truck route planning method integrating dynamic traffic flow and differentiated toll policies according to claim 1, characterized in that, The step of projecting the truck toll coefficients from the toll strategy configuration set onto the road network impedance distribution map in a time-dimensional alignment process, so that each road segment unit carries the road segment impedance value and the corresponding truck toll coefficient at each analysis step in the time series, forms a time-varying hybrid cost road network representation structure, including: The freight truck toll coefficient sequence corresponding to the specified road segment unit in the target road network and the toll coefficient switching time sequence associated with the freight truck toll coefficient sequence are read from the toll policy configuration set. The time interval boundary between two adjacent freight truck toll coefficients with different values ​​and intervals is defined by the switching time mark in the toll coefficient switching time sequence. Based on the switching time markers in the cost factor switching time sequence, a toll period boundary vector for the specified road segment unit is constructed, wherein each component of the toll period boundary vector represents the start and end time of a time period interval; Extract the time-refined road network impedance representation corresponding to the specified road segment unit from the road network impedance distribution state map. The time-refined road network impedance representation contains an impedance value sequence indexed by timestamps. Each impedance value in the impedance value sequence corresponds to a collection time or transition time. Starting from the beginning of the planned time interval, sampling time points is performed at a preset analysis step interval to generate a full-interval analysis time sequence. The time interval between two adjacent analysis times in the full-interval analysis time sequence is the analysis step interval. For each analysis time in the full-range analysis time sequence, the road segment impedance value corresponding to the timestamp of the analysis time is extracted from the impedance value sequence. At the same time, the truck toll fee coefficient associated with the analysis time is determined according to the time interval to which the analysis time belongs in the toll period boundary vector. The road segment impedance value and the truck toll fee coefficient are paired as the road segment mixed cost entry for the analysis time. Construct a hybrid cost time series vector corresponding to the specified road segment unit. Each component position of the hybrid cost time series vector stores a hybrid cost entry for the road segment at the analysis time. The hybrid cost entry for the road segment includes a road segment impedance value field and a truck passage cost coefficient field. The process involves traversing all road segment units of the target road network, reading the freight vehicle toll coefficient sequence and toll coefficient switching time sequence from the toll strategy configuration set for each road segment unit, and generating corresponding hybrid cost time series vectors to form a road network-level hybrid cost time series set. The hybrid cost time series vectors of each road segment unit in the road network-level hybrid cost time series set are then associated and organized according to the spatial topological connection relationship of the road segment units. For adjacent road segment units with upstream and downstream connections, the hybrid cost entries of the adjacent road segment units at the same analysis time are aligned and arranged in the time-varying dimension to generate a time-varying hybrid cost road network representation structure.

4. The truck route planning method integrating dynamic traffic flow and differentiated toll policies according to claim 1, characterized in that, The process involves performing a time-varying path search on the time-varying hybrid cost road network representation structure based on the given starting and ending points of the truck's journey. This time-varying path search process reads the segment impedance value and truck toll coefficient corresponding to the entry time of the segment while traversing adjacent segment units, and combines them to calculate a comprehensive path selection cost. The output includes a complete path segment sequence and the entry and exit times of each segment unit in the path segment sequence, including: The starting position of the truck's journey is mapped to the nearest network node in the time-varying hybrid cost road network representation structure as the starting node for path search, and the ending position of the truck's journey is mapped to the nearest network node in the time-varying hybrid cost road network representation structure as the ending node for path search; The earliest entry time of the path search starting node within the planned time interval is set as the path search starting time point, and a path tree with the path search starting node as the root node is created in the path search data structure. The root node of the path tree records the path search starting node identifier and the path search starting time point. Starting from the path search starting node, visit each neighbor node that is directly connected to the current node through a road segment unit in sequence. For the target neighbor node that is currently traversed, calculate the road segment entry time of the road segment unit traversed from the current node to the target neighbor node. Based on the entry time of the road segment, the road segment impedance value and truck toll coefficient of the corresponding road segment unit of the target neighbor node are searched in the time-varying hybrid cost road network representation structure at the entry time of the road segment. The found road segment impedance value is converted into a time cost component through a preset time value coefficient, and the truck toll coefficient is used as a monetary cost component. The time cost component and the monetary cost component are input into the comprehensive path selection cost fusion function. The comprehensive path selection cost fusion function performs a weighted summation of the time cost component and the monetary cost component to generate the comprehensive path selection cost of this visit, expressed in a unified monetary unit. The comprehensive path selection cost is added to the cumulative path cost in the path tree from the starting node of the path search to the target neighbor node to obtain the cumulative cost of the candidate path passing through the target neighbor node; When there are different arrival paths for the same target neighbor node, compare the cumulative cost of the candidate paths corresponding to the different arrival paths, retain the arrival path with the smaller cumulative cost of the candidate path, and update the parent node association relationship of the target neighbor node in the path tree as well as the corresponding road segment entry time and road segment exit time. The process of repeatedly visiting neighboring nodes from the current node and calculating the cumulative cost of candidate paths continues until the path search termination node is marked as having completed the search. This yields a path tree branch from the path search start node to the path search termination node. The path tree branch is then traced back from the path search termination node to the path search start node. The segment unit identifiers, entry times, and exit times of all road segment units along the path are extracted. These are then combined to generate a complete path segment sequence and the entry and exit times of each road segment unit in the path segment sequence.

5. The truck route planning method integrating dynamic traffic flow and differentiated toll policies according to claim 4, characterized in that, After obtaining the complete path segment sequence, the method further includes: Starting from the entry time corresponding to the first segment unit of the complete path segment sequence, the truck toll fee coefficient and segment impedance value corresponding to each segment unit in the complete path segment sequence are extracted sequentially. The total cost contribution of each road segment unit is calculated by using the freight truck toll coefficient and the geometric length parameter of each road segment unit. The total cost contribution of all road segment units in the complete route sequence is then added up item by item to obtain the total cost expenditure of the route. The total time consumption contribution of each road segment unit is calculated by using the road segment impedance value and the geometric length parameter of each road segment unit. The total time consumption contribution of all road segment units in the complete path road segment sequence is then added up item by item to obtain the total path travel time. Extract all differential toll segment unit identifiers with different rates in the target road network from the toll strategy configuration set, determine whether the complete path segment sequence contains differential toll segment units belonging to the differential toll segment unit identifiers, and when the complete path segment sequence contains differential toll segment units belonging to the differential toll segment unit identifiers, backtrack the entry time of the differential toll segment unit in the complete path segment sequence, and obtain the freight truck toll coefficient of the differential toll segment unit in the corresponding time-varying hybrid cost road network representation structure based on the entry time as the differential toll reference coefficient; The cost-saving substitution amount of the differentiated road segment is calculated using the differentiated cost reference coefficient and the geometric length parameter of the differentiated road segment unit. The cost-saving substitution amount of the differentiated road segment represents the change in the cost expenditure of the truck on the differentiated road segment unit relative to the scenario without differentiated tolling. The total cost expenditure of the route, the total travel time of the route, and the cost-saving substitution amount of the differentiated road segment are used as independent evaluation dimensions and combined into multi-dimensional evaluation information of route cost. The multi-dimensional evaluation information of route cost is then added to the truck route planning and scheduling information. The truck route planning and scheduling information carrying the multi-dimensional evaluation information of the path cost is sent to the target scheduling terminal, triggering the target scheduling terminal to simultaneously display the truck driving route guidance operation and the multi-dimensional evaluation information of the path cost.

6. The truck route planning method integrating dynamic traffic flow and differentiated toll policies according to claim 1, characterized in that, The method further includes: Pre-determine a set of restricted road segment units with truck passage time constraints from the target road network, and extract the start time and end time of the restricted period for each restricted road segment unit; In the process of constructing the time-varying hybrid cost road network representation structure, for each restricted road segment unit in the restricted road segment unit set, all analysis times within the restricted time period interval are marked as prohibited from passing in the hybrid cost time series vector corresponding to the restricted road segment unit; During the time-varying path search process of the time-varying hybrid cost road network representation structure, when traversing to the restricted road segment unit, the expected entry time of the road segment reaching the restricted road segment unit is read. Determine whether the expected entry time of the road segment falls within the restricted time period of the restricted road segment unit. When the expected entry time of the road segment falls within the restricted time period, assign a passage prohibition identifier to the restricted road segment unit and temporarily set its corresponding comprehensive path selection cost to an infinite amount of passage cost. When the expected entry time of the road segment does not fall within the restricted time period of the restricted road segment unit, the comprehensive route selection cost is calculated normally according to the road segment impedance value and truck toll coefficient corresponding to the expected entry time of the road segment in the time-varying hybrid cost road network representation structure. During the path search process, detour avoidance operations are performed on road segment units containing prohibition signs, so that the time-varying path search process automatically skips road segment units in the prohibition state and selects alternative connecting paths. After outputting the complete route segment sequence, the complete route segment sequence is reviewed for compliance based on the restricted road segment unit set. It is checked whether the entry time of each road segment unit in the complete route segment sequence overlaps with the restricted time period associated with that road segment unit. The complete route segment sequence that passes the compliance review is used as the route content in the truck route planning and scheduling information, and the prohibited passage status label of the restricted road segment unit set is added to the truck route planning and scheduling information.

7. The truck route planning method integrating dynamic traffic flow and differentiated toll policies according to claim 1, characterized in that, The method further includes: The historical roadside sensor datasets and historical tolling strategy configuration sets of multiple historical dates within the same planning time interval in the target road network are obtained to form a historical traffic scene sample set. For each set of historical roadside sensor datasets in the historical traffic scene sample set, a historical road network impedance distribution state map is generated by using the step of assigning time-varying impedance values ​​to all road segment units of the target road network using cross-sectional average traffic flow rate records and cross-sectional traffic flow space occupancy rate records. For each set of historical tolling strategy configurations in the historical traffic scenario sample set, a historical time-varying hybrid cost road network representation structure is generated by aligning and projecting the truck tolling coefficients in the tolling strategy configuration set to the road network impedance distribution state map in a time dimension. On each set of the historical time-varying hybrid cost road network representation structure, the time-varying path search process is performed for multiple different combinations of virtual truck trip start positions and virtual truck trip end positions to obtain multiple historical complete path segment sequences and the corresponding historical path total cost expenditure and historical path total trip time consumption. The historical complete route segment sequences of the same virtual truck trip start position and virtual truck trip end position under different historical dates are clustered according to the similarity of the route segment units to generate typical route clusters representing the corresponding start-end combination. For each typical route cluster, the entry time distribution range and exit time distribution range of each segment are extracted from the corresponding historical complete route segment sequence to generate the trip time band of the typical route cluster. In addition, for each typical route cluster, the cost distribution range and trip time distribution range are extracted from the corresponding historical total cost expenditure and historical total trip time to generate the cost distribution range of the typical route cluster. After obtaining the current complete path segment sequence by performing the time-varying path search process within the current planning cycle, the current complete path segment sequence is matched with the typical path cluster corresponding to the same combination of the truck's starting position and the truck's ending position to determine whether the current complete path segment sequence falls within the travel time band and cost distribution range of the typical path cluster. When the deviation direction between the current complete path segment sequence and the travel time band and cost distribution range of the typical path cluster indicates that the total path cost of the current complete path segment sequence is too high, historical complete path segment sequences are extracted from the typical path cluster as alternative path suggestions, and the alternative path suggestions are added to the truck route planning and scheduling information. The truck route planning and scheduling information with the added alternative path suggestions is then sent to the target scheduling terminal.

8. The truck route planning method integrating dynamic traffic flow and differentiated toll policies according to claim 1, characterized in that, The method further includes: A dedicated truck lane network is marked from the target road network. The dedicated truck lane network consists of dedicated road segment units that allow trucks to pass and nodes connecting the dedicated road segment units. The dedicated truck lane network connects the starting position and the ending position of the truck's journey with dedicated lane continuity. Extract the dedicated road segment hybrid cost entries corresponding to the dedicated road segment units in the dedicated truck lane network from the time-varying hybrid cost road network representation structure to form a dedicated lane cost substructure; The toll fee coefficients of the dedicated road segment units corresponding to the toll strategy configuration set are subjected to coefficient sensitivity classification processing, and the dedicated road segment units whose deviation from the preset benchmark toll coefficient exceeds the preset deviation threshold are marked as target toll road segment units. Before performing the time-varying path search process on the dedicated truck lane network, a cost coefficient copy adjustment is generated for each target toll road segment unit. The cost coefficient copy adjustment is obtained by increasing or decreasing the original truck toll coefficient of the target toll road segment unit. The cost coefficient copy of the target toll road segment unit is temporarily replaced with the original truck toll coefficient of the segment unit at the corresponding analysis time to generate an adjusted dedicated lane cost substructure. The time-varying path search process is performed on the adjusted dedicated lane cost substructure to obtain the adjusted complete path segment sequence and the corresponding adjusted total path cost. Based on the complete path segment sequences corresponding to the adjusted dedicated lane cost substructure and the original dedicated lane cost substructure, the differences between the two in terms of total path cost and total path travel time are compared to generate a cost coefficient sensitivity influence vector. The cost factor sensitivity influence vector is compared with the complete path segment sequence obtained by performing the time-varying path search processing on the time-varying hybrid cost road network representation structure. A list of road segment units sensitive to cost factor changes is extracted from the complete path segment sequence. The list of road segment units sensitive to cost factor changes and the cost factor sensitivity influence vector are combined into a cost sensitivity analysis result. The cost sensitivity analysis result is then appended to the truck route planning and scheduling information. The truck route planning and scheduling information carrying the cost sensitivity analysis results is sent to the target scheduling terminal, triggering the target scheduling terminal to simultaneously display the list of road segment units sensitive to cost coefficient changes when displaying the truck driving route guidance operation.

9. The truck route planning method integrating dynamic traffic flow and differentiated toll policies according to claim 1, characterized in that, After generating the truck route planning and scheduling information, the method further includes: Incremental roadside sensor data and incremental toll policy update records are collected from the roadside sensor dataset and toll policy configuration set of the target road network at a preset real-time refresh cycle. The time-varying impedance assignment processing is performed on the newly added cross-section average traffic flow rate record and newly added cross-section traffic flow space occupancy record in the incremental roadside sensor data to generate an incremental road network impedance distribution update segment. The time dimension alignment projection process is performed on the newly added truck toll fee coefficient and the switching time sequence of the newly added toll fee coefficient in the incremental toll policy update record to generate an incremental hybrid cost update fragment. The affected road segment units and time segment ranges in the time-varying hybrid cost road network representation structure are locally replaced by the incremental road network impedance distribution update fragment and the incremental hybrid cost update fragment to obtain a rolling updated time-varying hybrid cost road network representation structure. The actual position of the truck at the current moment is mapped to the rolling updated time-varying hybrid cost road network representation structure as the replanning start node, and the end position of the truck's journey is used as the replanning end node; The time-varying path search process is performed on the rolling updated time-varying hybrid cost road network representation structure to generate a replanned complete path segment sequence and the replanning entry time and replanning exit time of each segment unit in the replanned complete path segment sequence. The replanned complete path segment sequence is compared with the original complete path segment sequence to calculate the degree of path deviation. The consistency between the replanned complete path segment sequence and the original complete path segment sequence in terms of segment unit composition is determined. When the consistency indicates that the path has deviated, the newly added and removed segment units in the replanned complete path segment sequence are extracted to generate a list of path change differences. The replanned complete route segment sequence, the replanned entry time, the replanned departure time, and the route change difference list are combined to form updated route planning and scheduling information, which is then sent to the target scheduling terminal to trigger dynamic adjustment guidance of the truck's driving route.

10. A truck route planning system integrating dynamic traffic flow and differentiated toll policies, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the truck routing method according to any one of claims 1 to 9 by executing the machine-executable instructions.

Images