CAV special lane layout optimization method based on travel time reliability

Abstract

The invention discloses a travel time reliability-based CAV special lane layout optimization method. The method comprises the following steps of: constructing an expanded road network comprising a conventional road section and a virtual road section; calculating the traffic capacity of the conventional road section and the virtual road section based on the expanded road network according to the vehicle following mode and the corresponding safe time headway; establishing a mixed traffic flow distribution model based on the traffic capacity, and solving the mixed traffic flow distribution model to obtain road network balanced flow distribution; obtaining a travel time sample set of the road sections and the paths through Monte Carlo random sampling according to the balanced flow distribution; calculating a road network buffer time index based on the travel time sample set, wherein the buffer time index represents road network travel time reliability; the layout number of the CAV special lanes on the virtual road section is taken as a decision variable, the minimum road network total travel time and the minimum road network total buffer time index are taken as optimization objectives, a CAV special lane layout optimization model is established, and the model is used for outputting a CAV special lane layout optimization scheme.

Description

Technical Field

[0001] This invention belongs to the field of smart Internet of Things technology, and in particular relates to a method for optimizing the layout of CAV dedicated lanes based on travel time reliability. Background Technology

[0002] With the rapid development of intelligent connected vehicle technology, the coexistence and mixed driving of connected and autonomous vehicles (CAVs) and human-driven vehicles (HVs) in traffic networks will become a typical traffic scenario.

[0003] In mixed-traffic scenarios, different vehicle combinations (such as HV following HV, CAV following CAV, etc.) correspond to different safe headway distances, and the order in which vehicles are arranged on the road segment is completely random. This makes the average headway distance on a road segment in mixed-traffic scenarios a random variable related to the spatial distribution of different types of vehicles on the road segment. Since the average headway distance on a road segment directly determines the traffic capacity of the road segment, the traffic capacity of the road segment in mixed-traffic scenarios is also a random variable. The size of the traffic capacity of a road segment directly affects the travel time of vehicles on that road segment. Therefore, in mixed-traffic scenarios, due to the randomness of the traffic capacity of a road segment, the travel time of a road segment will exhibit huge fluctuations, causing changes in the reliability of the travel time of that road segment, which will further transmit to the entire road network and ultimately affect the reliability of the travel time of the entire road network. However, there is currently a lack of a method that can accurately quantify and evaluate the reliability of the travel time of the road network in mixed-traffic scenarios. Furthermore, from the perspective of traffic planning and management, how to improve the reliability of road network travel time through the optimized layout of CAVs is also a major challenge. To this end, this invention proposes a CAV dedicated lane layout method that can take into account both "minimum total travel time of the road network" and "optimal total travel time reliability of the road network". This method can provide traffic managers with scientific CAV dedicated lane layout decisions based on different travel efficiency needs and different travel time reliability preferences. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a CAV lane layout optimization method based on travel time reliability, which can provide traffic managers with scientific CAV lane layout decisions based on different travel efficiency needs and different travel time reliability preferences.

[0005] To achieve the above objectives, this invention provides a CAV (Cross-Traffic Access Vehicle) lane layout optimization method based on travel time reliability, comprising: Construct an expanded road network that includes both conventional and virtual road sections; Based on the expanded road network, the traffic capacity of regular road sections and virtual road sections is calculated according to the vehicle following mode and the corresponding safe headway. A mixed traffic flow allocation model is established based on the aforementioned traffic capacity, and the balanced traffic flow distribution of the road network is obtained by solving the mixed traffic flow allocation model. Based on the balanced traffic distribution, a set of travel time samples for road segments and routes is obtained through Monte Carlo random sampling; The road network buffer time index is calculated based on the travel time sample set, and the buffer time index characterizes the reliability of road network travel time. Using the number of dedicated CAV lanes on virtual road segments as the decision variable, and minimizing the total travel time and the total buffer time exponent of the road network as the optimization objectives, a CAV lane layout optimization model is established. The CAV lane layout optimization model is used to output a set of CAV lane layout optimization schemes.

[0006] Optionally, constructing the extended road network includes: For each segment of the initial road network, a virtual segment is added, and the initial segment is designated as a regular segment. The sum of the number of lanes in the regular segment and the virtual segment is equal to the total number of lanes in the corresponding segment of the initial road network. The regular segment allows connected autonomous vehicles and human-driven vehicles to travel together, while the virtual segment only allows connected autonomous vehicles to travel.

[0007] Optionally, the vehicle following mode includes: On the regular road sections, there are five modes: human-driven vehicles following human-driven vehicles, human-driven vehicles following connected autonomous vehicles, connected autonomous vehicles following human-driven vehicles, connected autonomous vehicles following connected autonomous vehicles in the same platoon, and connected autonomous vehicles following connected autonomous vehicles in different platoons. The virtual road segment is divided into two modes: connected autonomous vehicles following connected autonomous vehicles in the same platoon, and connected autonomous vehicles following connected autonomous vehicles in different platoons.

[0008] Optionally, the traffic capacity of the conventional road section is: ; in, To ensure the traffic capacity of regular road sections, This refers to the number of lanes on a regular road section. This represents the average headway on a typical road section. , , , , Different vehicle following modes for regular road sections , , , for , , , , Corresponding to different safe headway distances, To expand the collection of regular road sections in the road network; The traffic capacity of the virtual road segment is: ; in, For the traffic capacity of virtual road segments, This represents the average headway on the virtual road segment. , Different vehicle following modes for virtual road segments. , for , Corresponding to different safe headway distances, This refers to the number of lanes on a virtual road segment. To expand the set of virtual road segments in the road network.

[0009] Optionally, based on the traffic capacity, a mixed traffic flow assignment model is established, and solving the mixed traffic flow assignment model to obtain the road network equilibrium traffic distribution includes: Based on the aforementioned capacity, a variational inequality problem is constructed, which includes a path cost function, a flow conservation constraint, and a variable nonnegativity constraint. The path cost function is formed by aggregating the travel times of road segments, and the travel times of road segments are calculated using the BPR function based on the flow and capacity of the road segments. The variational inequality problem is solved using an iterative algorithm based on path exchange until the convergence condition is met, thereby obtaining the equilibrium flow distribution.

[0010] Optionally, based on the balanced traffic distribution, the travel time sample set of road segments and routes obtained through Monte Carlo random sampling includes: S1. Based on the balanced traffic distribution, generate a random permutation sequence of connected autonomous vehicles and human-driven vehicles for each road segment; S2. Identify the following pattern between vehicles in the random sequence and assign a corresponding safe headway. Calculate the average headway of the road segment based on the safe headway of all vehicle pairs. S3. Calculate the road segment capacity based on the average headway of the road segment, and calculate the travel time of the road segment using the BPR function; S4. Calculate the route travel time based on the road segments included in the route and the travel time of each road segment. S5, Repeat S1-S4 Next, retrieve each Construct a set of route travel times from a set of different travel time samples.

[0011] Optionally, calculating the road network buffer time index based on the travel time sample set includes: Extract the 95th quantile from the sample set of path travel times as the planned travel time, and calculate the average value of the sample set of path travel times as the average travel time. The route buffer time index is obtained by dividing the difference between the planned travel time and the average travel time by the average travel time. The buffer time index of origin-destination pairs is obtained by taking a flow-weighted average of all path buffer time indices between origin-destination pairs in the road network. The total buffer time index of the road network is obtained by taking a flow-weighted average of the buffer time indexes for all origin and destination points.

[0012] Optionally, the constraints of the CAV dedicated lane layout optimization model include: lane resource conservation constraints, traffic flow conservation constraints, user equilibrium constraints, and generalized travel cost constraints.

[0013] Optionally, the CAV dedicated lane layout optimization model is used to output a set of CAV dedicated lane layout optimization schemes, including: solving the CAV dedicated lane layout optimization model using a fast elite multi-objective genetic algorithm; The CAV dedicated lane layout optimization model is solved using a fast elite multi-objective genetic algorithm, including: S1. Encode the virtual road segment layout scheme with integers and randomly generate an initial population; S2. The offspring population is generated through selection operations based on non-dominated sorting and crowding distance, simulated binary crossover operations, and polynomial mutation operations. S3. After merging the parent and offspring populations, perform rapid non-dominated sorting and select individuals to form a new generation population based on sorting level and crowding distance. S4. Iterate through S1-S3 until the maximum number of generations is reached, and output the solutions in the non-dominated levels as the set of CAV dedicated lane layout optimization solutions.

[0014] Compared with the prior art, the present invention has the following advantages and technical effects: This invention is the first to combine the quantitative assessment of travel reliability in mixed-traffic road networks with a CAV (Caravan Access Vehicle) lane layout optimization system, constructing a complete system from evaluation to optimization. By constructing a mixed-traffic flow distribution model considering five car-following patterns and employing Monte Carlo sampling with complete random sampling, this invention achieves a quantitative assessment of road network travel time reliability. Furthermore, by constructing an optimization model aimed at minimizing total road network travel time and maximizing total road network travel time reliability, it achieves synergistic optimization of road network travel efficiency and travel time reliability. This invention provides traffic managers with a complete methodology from reliability assessment to optimization scheme generation, supporting them in making scientific trade-offs and decisions between travel efficiency and travel time reliability. Attached Figure Description

[0015] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of a CAV dedicated lane layout optimization method based on travel time reliability according to an embodiment of the present invention; Figure 2 This is an iterative curve diagram of Nguyen-Dupuis road network optimization according to an embodiment of the present invention. Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0018] This embodiment proposes a CAV dedicated lane layout optimization method based on travel time reliability, such as... Figure 1 As shown, the specific steps include: Construct an expanded road network that includes both conventional and virtual road sections; Based on the expanded road network, the traffic capacity of regular road sections and virtual road sections is calculated according to the vehicle following mode and the corresponding safe headway. A mixed traffic flow allocation model is established based on the aforementioned traffic capacity, and the balanced traffic flow distribution of the road network is obtained by solving the mixed traffic flow allocation model. Based on the balanced traffic distribution, a set of travel time samples for road segments and routes is obtained through Monte Carlo random sampling; The road network buffer time index is calculated based on the travel time sample set, and the buffer time index characterizes the reliability of road network travel time. Using the number of dedicated CAV lanes on virtual road segments as the decision variable, and minimizing the total travel time and the total buffer time exponent of the road network as the optimization objectives, a CAV lane layout optimization model is established. The CAV lane layout optimization model is used to output a set of CAV lane layout optimization schemes.

[0019] Specifically, this embodiment takes "minimum total travel time of the road network" and "optimal reliability of total travel time of the road network" as optimization objectives, establishes a CAV dedicated lane layout optimization model and its solution algorithm, and provides traffic planners with a set of CAV dedicated lane layout optimization schemes based on different travel efficiencies and different reliability preferences.

[0020] Furthermore, constructing the extended road network includes: For each segment of the initial road network, a virtual segment is added, and the initial segment is designated as a regular segment. The sum of the number of lanes in the regular segment and the virtual segment is equal to the total number of lanes in the corresponding segment of the initial road network. The regular segment allows connected autonomous vehicles and human-driven vehicles to travel together, while the virtual segment only allows connected autonomous vehicles to travel.

[0021] Specifically, constructing and expanding the road network: make Indicates the initial road network. Represents the set of nodes in the road network. Represents the set of road segments in a road network. and These are the sets of starting points and ending points in the road network. , , and They represent a starting point and an ending point, respectively. , . Indicates the starting point is The destination is The section of road, .

[0022] Constructing and expanding the road network For the initial road network For each road segment in the system, a virtual road segment is added, where the initial road segment is denoted as the regular road segment. Virtual road segments are denoted as virtual road segments. ,satisfy .

[0023] Furthermore, the vehicle following mode includes: On the regular road sections, there are five modes: human-driven vehicles following human-driven vehicles, human-driven vehicles following connected autonomous vehicles, connected autonomous vehicles following human-driven vehicles, connected autonomous vehicles following connected autonomous vehicles in the same platoon, and connected autonomous vehicles following connected autonomous vehicles in different platoons. The virtual road segment is divided into two modes: connected autonomous vehicles following connected autonomous vehicles in the same platoon, and connected autonomous vehicles following connected autonomous vehicles in different platoons.

[0024] Furthermore, based on the aforementioned traffic capacity, a mixed traffic flow allocation model is established. Solving the mixed traffic flow allocation model to obtain the equilibrium traffic distribution of the road network includes: Based on the aforementioned capacity, a variational inequality problem is constructed, which includes a path cost function, a flow conservation constraint, and a variable nonnegativity constraint. The path cost function is formed by aggregating the travel times of road segments, and the travel times of road segments are calculated using the BPR function based on the flow and capacity of the road segments. The variational inequality problem is solved using an iterative algorithm based on path exchange until the convergence condition is met, thereby obtaining the equilibrium flow distribution.

[0025] Specifically, (2) Establish a traffic flow allocation model for mixed-traffic scenarios: This embodiment constructs a traffic flow allocation model for mixed-traffic scenarios to obtain the spatial distribution of mixed-traffic flow on the road network. For a known extended road network, the model first establishes a functional expression for the road segment capacity and travel time based on vehicle car-following patterns. It then describes the path selection behavior of HV and CAV according to the user equilibrium criterion. Finally, by constructing a mixed-traffic flow allocation model in the form of a variational inequality problem, the model is solved using a path transformation algorithm to obtain the HV and CAV flow on each road segment and path when the road network reaches equilibrium.

[0026] 1) Calculation of travel costs for road segments and routes: Define the vehicle following modes and corresponding safe headway in mixed-traffic scenarios. Considering the maximum size limit of CAV platoons, use... This indicates the maximum number of vehicles allowed in a CAV fleet, with a formation size reaching [number missing]. At this time, a separate convoy of CAVs is required. On regular roads, based on whether the vehicles in front and behind belong to the same convoy, five vehicle following modes are defined: one HV following another HV, one HV following one CAV, one CAV following one HV, one CAV following one CAV within the convoy, and one CAV following one CAV outside the convoy. The different safe headway distances corresponding to the above following modes are expressed as follows: , , , On the virtual road segment, two vehicle following modes are defined: one CAV follows another CAV within a platoon, and one CAV follows another CAV outside a platoon. The corresponding safe headway distances are expressed as follows: , .

[0027] Record any section of the expanded road network The probability that the previous vehicle was a CAV is The calculation formula is as follows: (1); in, Indicates any segment of the extended road network The total number of CAVs on the platform Indicates any segment of the extended road network The total number of HV values.

[0028] Calculate the percentage of different car-following modes. Record the percentages of the five car-following modes as follows: , , , , When vehicles on any segment of the expanded road network are completely randomly distributed and The proportions of these five car-following modes can be calculated using equations (2) to (6): (2); (3); (4); (5); (6); when hour, , , Since both are 0, the limit can be derived using L'Hôpital's limit rule. and The calculation formula is as follows: (7); (8); For the initial road network any section of the road Record the total number of lanes as Expand the road network In the middle, the number of lanes on a regular road section is recorded as follows: The number of lanes on the virtual road segment is The sum of the number of lanes in the regular road segment and the virtual road segment should equal the total number of lanes in the corresponding road segment in the initial road network, that is: (9); (10); Based on different safe headway distances and their corresponding percentages, the following can be calculated for regular road sections. The average headway on the road is: (11); regular road sections The traffic capacity of the road segment Number of lanes and road sections The product of the reciprocal of the average headway, i.e.: (12); Similarly, virtual road segments The average headway and traffic capacity can be calculated using equations (13) and (14), respectively: (13); (14); regular road sections Travel time can be calculated using the Bureau of Public Roads (BPR) function: (15); in, and These represent regular road sections. HV and CAV traffic on the device. Indicates road segment The free-flow time of all vehicles.

[0029] Similarly, virtual road segments The CAV travel time function is: (16); in, Virtual road segment Traveling on the section of road leading to CAV; Indicates virtual road segment CAV traffic on the platform.

[0030] HVs are not allowed to travel on virtual road segments. Therefore, the travel time for HVs on virtual road segments is set to be infinite, as expressed by the following formula: (17); in, Indicates virtual road segment Travel time on HV-dedicated lanes for HV vehicles; It is a sufficiently large normal quantity.

[0031] The route cost of a vehicle is the sum of the travel times of all the road segments it includes. In expanding the road network... In the middle, the origin and destination points are defined as a pair. The set of feasible paths is Then the model In the path The travel cost function is: (18); in, According to vehicle model and road sections The types are taken respectively. , or ; It is to determine the road segment Is it on the path? The constraint on the value is 1 if it is true, otherwise it is 0.

[0032] 2) Flow constraints: We assume that all HV and CAV travelers are rational and follow the user equilibrium criterion.

[0033] The sum of the traffic flow of a certain type on all paths between the origin and destination should equal the travel demand of that type of vehicle at the origin and destination: (19); in, Indicates vehicle model path Traffic flow on the road Indicates origin-end point pair Interval models The demand for travel.

[0034] The HV and CAV traffic on each path should be non-negative, that is: (20); Section The traffic flow on a given road segment is equal to the sum of the traffic flows of all routes passing through that segment, i.e.: (twenty one); in, Indicates road segment Up vehicle The quantity.

[0035] 3) User equilibrium modeling: make Indicates vehicle model The column vector of travel costs, since travel time on a road segment is affected by the traffic flow of all vehicle types on that segment, therefore, It is about the flow vector of all paths. The function. Let Indicates origin-end point pair Interval models The minimum cost of travel along the shortest path.

[0036] The user equilibrium condition can be expressed as follows: for any vehicle model and origin and destination pairs The cost of the used path is equal to and equal to the minimum cost, and the cost of the unused path is not lower than the minimum cost, that is: (twenty two); in, This represents the path flow vector under equilibrium conditions. Indicates the vehicle model under equilibrium conditions. In the path Traffic on the internet.

[0037] The user equilibrium problem described above can be transformed into the following variational inequality problem: (twenty three); Among them, feasible region It is a convex set defined by constraints (19) and (20): (twenty four); 4) Solution algorithm based on path switching: To solve the variational inequality problem in 3), this embodiment uses a solution algorithm based on path exchange. This algorithm iteratively transfers traffic from high-cost paths to low-cost paths in the known set of paths, gradually approaching the user equilibrium state.

[0038] i. Generate a set of feasible paths: The algorithm first needs to expand the road network. Each start and end point pair Generate a set of feasible paths containing a finite number of paths. This invention employs the K-shortest path algorithm for path enumeration: Step 1: Find the initial road network From the starting point To the finish line The first path with the lowest travel cost is denoted as . ; Step 2: For the initial road network Each path that has been found It attempts to deviate from the original path at a certain intermediate node, explores new adjacent nodes, and thus generates multiple candidate deviation paths; Step 3: From the initial road network Among all candidate paths, select the one with the lowest travel cost and no duplicates, add it to the set of found paths, and use it as the next shortest path. Step 4: Repeat steps 2 and 3 until the predetermined quantity is found. These paths constitute the initial road network. feasible path set .

[0039] Step 5: Initial road network feasible path set For each regular path in the network, find its application in the expanded road network. The virtual paths (composed of virtual road segments) that correspond one-to-one with the nodes form an extended road network that includes both regular paths and virtual paths. feasible path set .

[0040] ii. Traffic initialization and iterative updates: Step 1: Traffic initialization. Set the number of iterations. Pair the origin and destination points The number of CAV vehicles required by each location is evenly distributed across the expanded road network. feasible path set On all paths, including regular paths and virtual paths, the origin and destination pairs will be... The number of HV vehicles required by each location is evenly distributed across the expanded road network. feasible path set On the regular paths in the network, obtain the initial path traffic for all paths in the extended road network. ,in .

[0041] Step 2: Iterative update of path traffic.

[0042] Step 2.1: Set path traffic update rules. (In the...) In this iteration, the path traffic is updated according to the following rules: (25); in, It is the first The path flow vector for the next iteration. It is the flow update direction vector, where each component... Represents the model In the path The direction and intensity of traffic updates. It is the first The step size of each iteration is used to control the update magnitude.

[0043] Step 2.2: Determine the traffic update direction. Update direction The calculation embodies the idea of ​​"path swapping": the current path is swapped. Some traffic is redirected to other, lower-cost paths within the same origin-destination pair. The calculation formula is as follows: (26); in, Indicates origin-end point pair medium-sized models The actual set of usable paths (for HV, , only contains For CAV, the section of road ). and They represent the first and second parts respectively. Path and the Vehicle type on the route Travel costs. .

[0044] Step 2.3: Set the traffic update step size. Step size An adaptive strategy is employed to determine this, balancing convergence speed and stability: (27); in, It is the maximum magnitude estimate of the current iteration update direction. It is a scaling factor that adjusts over time.

[0045] (28); (29); in, and As a preset constant, when the norm of the update direction increases (the algorithm may diverge), make Increase rapidly, thereby reducing the step size. With a stable algorithm; when the update direction norm decreases (the algorithm tends to converge), make Increase the step size slowly and moderately to accelerate convergence.

[0046] Step 2.4: Set the convergence metric and its threshold. The iterative process continues until the preset convergence condition is met. Define the convergence metric. A small positive number is given to indicate the degree of deviation between the current flow and the equilibrium condition. , as the convergence value. The calculation formula is as follows: (30); in, It is the vehicle model v in the current iteration at the origin and destination points. The minimum path cost between.

[0047] Step 3: If If the algorithm has converged, it stops iterating and outputs the current path traffic. As an equilibrium path flow An approximate solution; otherwise, let Return to step 2 and continue iterating.

[0048] Furthermore, based on the aforementioned balanced traffic distribution, the travel time sample set for road segments and routes obtained through Monte Carlo random sampling includes: S1. Based on the balanced traffic distribution, generate a random permutation sequence of connected autonomous vehicles and human-driven vehicles for each road segment; S2. Identify the following pattern between vehicles in the random sequence and assign a corresponding safe headway. Calculate the average headway of the road segment based on the safe headway of all vehicle pairs. S3. Calculate the road segment capacity based on the average headway of the road segment, and calculate the travel time of the road segment using the BPR function; S4. Calculate the route travel time based on the road segments included in the route and the travel time of each road segment. S5, Repeat S1-S4 Next, retrieve each Construct a set of route travel times from a set of different travel time samples.

[0049] Specifically, (3) reliability indicators and calculation of mixed-traffic road networks: In the traffic flow distribution model in (2), setting the number of vehicles in all CAV lanes to 0 allows us to obtain the balanced traffic flow distributed on each road segment of a mixed-traffic network without CAV lanes. and Monte Carlo sampling simulates the spatial distribution of different types of vehicles on road segments, allowing us to obtain travel time samples for vehicles on road segments and routes. The specific steps of Monte Carlo sampling are as follows: Step 1: Generate vehicle sequence: For each road segment superior CAV and The vehicles are arranged in a completely random manner to form a length of The vehicle sequence.

[0050] Step 2: Car-following pattern recognition and headway calculation: Traverse the vehicle sequence from Step 1, determine the car-following pattern between each pair of vehicles, assign a corresponding safe headway, add up the safe headway of all vehicle pairs on this road segment, and then divide by the total number of vehicle pairs to calculate the average headway of this road segment. .

[0051] Step 3: Calculation of road segment capacity and travel time: based on the average headway of the road segment Calculate the traffic capacity of the road segment Then, the BPR function is used to calculate the travel time for the road segment.

[0052] Step 4: Path travel time aggregation: The travel time for each path can be calculated according to formula (18).

[0053] Step 5: Repeat the above steps Next, obtain the path for each... Construct a set of route travel times from a set of different travel time samples.

[0054] Furthermore, calculating the road network buffer time index based on the aforementioned path travel time sample set includes: Extract the 95th quantile from the sample set of path travel times as the planned travel time, and calculate the average value of the sample set of path travel times as the average travel time. The route buffer time index is obtained by dividing the difference between the planned travel time and the average travel time by the average travel time. The buffer time index of the origin-destination pair is obtained by taking a flow-weighted average of all path buffer time indices between the origin and destination points. The total buffer time index of the road network is obtained by taking a flow-weighted average of the buffer time indexes for all origin and destination points.

[0055] Specifically, the buffer time index is chosen as the evaluation index for travel time reliability of the road network because it overcomes the dimensional dependence problem of statistical indicators, and its calculation does not require travel times to follow a specific distribution. It is suitable for complex data environments in mixed traffic scenarios, and the smaller the buffer time index, the higher the travel time reliability. Path The formula for calculating the upper buffer time exponent is as follows: (31); in, It is a path The travel time of the route ranked 95th in the set of travel times arranged in ascending order; It is a path Path in the set of travel times The average travel time.

[0056] Origin and end point pairs The buffer time index can be determined by the start and end points. The exponentially weighted average of the buffer times for all paths is calculated as follows: (32); Similarly, the total buffer time index of the road network can be calculated by weighted average of the buffer time indices of all origin-destination pairs in the road network: (33); in, For the set of origin-end point pairs, For origin and destination pairs Traffic demand between areas.

[0057] In summary, once the buffer time index of the road network is calculated, the evaluation of the reliability of road network travel in mixed traffic scenarios is completed.

[0058] Furthermore, the constraints of the CAV dedicated lane layout optimization model include: lane resource conservation constraints, flow conservation constraints, user equilibrium constraints, and generalized travel cost constraints.

[0059] Furthermore, the CAV dedicated lane layout optimization model is used to output a set of CAV dedicated lane layout optimization schemes, including: solving the CAV dedicated lane layout optimization model using a fast elite multi-objective genetic algorithm; The CAV dedicated lane layout optimization model is solved using a fast elite multi-objective genetic algorithm, including: S1. Encode the virtual road segment layout scheme with integers and randomly generate an initial population; S2. The offspring population is generated through selection operations based on non-dominated sorting and crowding distance, simulated binary crossover operations, and polynomial mutation operations. S3. After merging the parent and offspring populations, perform rapid non-dominated sorting and select individuals to form a new generation population based on sorting level and crowding distance. S4. Iterate through S1-S3 until the maximum number of generations is reached, and output the solutions in the non-dominated levels as the set of CAV dedicated lane layout optimization solutions.

[0060] Specifically, a CAV dedicated lane optimization model was constructed with the objectives of minimizing the total travel time of the road network and maximizing the reliability of the total travel time of the road network. A fast elite multi-objective genetic algorithm was used to solve the model. This achieved optimization of the reliability of the total travel time of the road network under mixed-traffic scenarios.

[0061] (1) Optimization model: This example constructs a CAV dedicated lane layout optimization model, aiming to maximize the reliability of the road network travel time while minimizing the total travel time of the road network by optimizing the CAV dedicated lane layout.

[0062] 1) Decision variables: The decision variable in the CAV dedicated lane layout optimization model is the CAV dedicated lane layout scheme, defined as: (34); in, Virtual road segment The number of lanes for the CAV dedicated lanes set above is a non-negative integer.

[0063] 2) Objective function: Objective function 1: (35); in, Let be the decision variable, representing the CAV dedicated lane layout optimization scheme; For the layout plan of the dedicated lane Total travel time on the road network; Indicates road segment Up vehicle Travel time for the route; Demonstration section Up vehicle Traffic flow on the road section.

[0064] Objective function 2: (36); in, Under the dedicated lane layout scheme The total buffer time index of the road network is the highest reliability of the total travel time of the road network. The smaller the total buffer time index of the road network, the higher the reliability of the total travel time of the road network. For origin and destination pairs Reliability of travel time.

[0065] 3) Constraints: The optimized model must satisfy the following four types of constraints: i. Lane resource constraints: including lane number conservation constraint (9) and non-negative integer constraint (10). ii. Generalized travel cost constraints: Equations (1) and (15)-(17); iii. Flow conservation constraints: Equations (19)-(21); iv. User equilibrium constraints: Equations (22)-(24); (2) Objective function calculation process: The objective function calculation process in this embodiment (1) is closely dependent on the established mixed traffic flow assignment model and reliability evaluation method. The specific calculation process is as follows: Step 1: Input and Initialization. Input a candidate CAV dedicated lane layout optimization scheme. The number of lanes in the corresponding conventional road segment is determined according to the lane number conservation constraint (9) and the non-negative integer constraint (10). The expanded road network is then initialized. and all model parameters.

[0066] Step 2: Implement mixed traffic flow assignment. (Based on the deployment plan) Defined extended road network Given the origin and destination requirements as input, the system calls the mixed traffic flow allocation model and its solution algorithm, and finally outputs the segment traffic flow distribution and path traffic flow distribution when the road network reaches an equilibrium state.

[0067] Step 3: Calculate the total travel time of the road network. Substitute the equilibrium traffic flow of each road segment obtained in Step 2 into Equations (15)-(17) to obtain the equilibrium travel time of each road segment. Then, according to objective function 1, multiply the travel time of each road segment by the equilibrium traffic flow of the road segment and sum them to obtain the total travel time of the road network under this layout scheme.

[0068] Step 4: Calculate the overall reliability of the road network.

[0069] Step 4.1: Monte Carlo Sampling: Using the balanced CAV and HV traffic volumes obtained in Step 2 as input, perform the Monte Carlo sampling step. Through large-scale random simulation of vehicle arrangement on the road segment, sample each path between each origin-destination pair. Generate a containing The set of travel times for each sample.

[0070] Step 4.2: Calculate each origin-end point pair according to equation (32). The weighted average buffer time exponent; the origin and destination points are paired. The weighted average buffer time index and its origin-end point demand are substituted into objective function 2 to calculate the overall reliability index of the road network.

[0071] Step 5: Output the deployment plan The two objective function values ​​are given below.

[0072] (3) Fast Elite Multi-Objective Genetic Solution Algorithm: The model in (1) is a nonlinear integer bi-objective optimization problem. In this embodiment, a fast elite multi-objective genetic algorithm is used to solve it. The specific steps are as follows: Step 1: Initialization and Propagation. Randomly generate I initial CAV dedicated lane layout schemes. The number of CAV dedicated lanes on each road segment is identified using integer codes. Perform genetic operations: generate offspring populations based on non-dominated sorting and crowding distance selection, simulated binary crossover, and polynomial mutation. .Will and They merged to form a joint population of size 2I. .

[0073] Step 2: Non-dominated ranking and elite preservation. For joint populations A fast non-dominated ranking system with dual objectives (minimizing total travel time and maximizing the reliability of total travel time) is implemented. All individuals in the population are divided into multiple non-dominated levels (non-dominated level 1, non-dominated level 2, ..., where non-dominated level 1 is the optimal non-dominated level) according to Pareto dominance. The congestion distance in the objective space is calculated for individuals within the same level. Based on level priority and congestion distance, I optimal and diverse layout schemes are selected to form a new generation of the population. .

[0074] Step 3: Iterative Evolution. (Regarding...) Repeat steps 1 and 2 until the maximum number of generations or Pareto solution set converges. Finally, output the solutions in non-dominated level 1 as the Pareto optimal CAV dedicated lane layout scheme set.

[0075] Case Analysis: To verify the feasibility of the travel reliability evaluation and optimization method for intelligent connected mixed traffic networks proposed in this invention, this embodiment selects the classic Nguyen-Dupuis road network as a case for simulation analysis.

[0076] (1) Parameter settings: The safe headway values ​​for the five car-following modes are as follows: , , , , The maximum platoon length L of the CAV is 6, the CAV penetration rate is 0.5, and the undetermined coefficients in the Bureau of Public Roads (BPR) function are... , Taking values ​​of 0.15 and 4 respectively, the Nguyen-Dupuis road network has a total of four origin-end point pairs, namely: , , , Set the start and end point requirements to 9000, 8500, 8600, and 9200 respectively, and the convergence value... .

[0077] (2) Reliability evaluation of mixed-traffic road network: First, the "Travel Reliability Evaluation Method for Intelligent Connected Mixed-Traffic Networks" is applied to a network without any dedicated CAV lanes ( The initial mixed-traffic network is evaluated. The specific operation steps are as follows: Step 1: Call the mixed traffic flow distribution model and its solution algorithm to calculate the equilibrium flow of HV and CAV on each road segment when the road network reaches user equilibrium. and .

[0078] Step 2: Based on the above balanced flow, strictly follow the Monte Carlo sampling procedure and set the number of simulations. This generates a large number of travel time samples for each road segment and route.

[0079] Step 3: Calculation of reliability index: Based on equations (31)-(33), the buffer time index of each path is calculated using the path travel time sample, and then the total buffer time index of each origin-destination pair and the entire road network is obtained by weighting the flow.

[0080] Through the complete process described above, the effective paths and their buffer time indices between each origin-destination pair in the initial road network (without CAV dedicated lanes) were obtained, as shown in Table 1. Simultaneously, the total buffer time index for the entire road network was calculated to be 0.2802. A higher buffer time index indicates more significant fluctuations in travel time and lower reliability. Travelers can use this value to compare the reliability of various paths and make rational judgments.

[0081] Table 1 (2) Optimization and Result Analysis of CAV Dedicated Lane: The constructed "Optimization Model for CAV Dedicated Lane Layout Oriented to Efficiency and Reliability" and its fast elite multi-objective genetic solution algorithm were used for optimization. The algorithm parameters were set as follows: initial population size I... The maximum number of generations is 80, the crossover probability is 0.9, and the mutation probability is 0.1.

[0082] After the algorithm was run, the Pareto front iterative curve of the Nguyen-Dupuis road network was obtained (e.g., Figure 2As shown in Figure 2, this figure illustrates the convergence and distribution of the Pareto front during the evolutionary process. The vertical axis represents the total buffer time exponent of the road network; a smaller value indicates higher travel time reliability. The horizontal axis represents the total travel time of the road network; a smaller value indicates higher overall travel efficiency. The curves of different colors in the figure represent the Pareto fronts obtained by the multi-objective genetic algorithm at different iteration numbers when solving the CAV lane layout optimization problem. Each point on each curve represents a non-dominated solution found at that iteration number, which is also a CAV lane layout scheme. Finally, a set of non-dominated Pareto-optimal CAV lane layout optimization schemes at the 80th iteration is output, as shown in Table 2. The arrays under the CAV lane layout optimization schemes represent the number of CAV lanes set on each virtual road segment in the extended road network. Each array corresponds to a layout scheme of a set of virtual road segments, and the total travel time and total buffer time exponent of the road network under this scheme are given.

[0083] Table 2 from Figure 2 As clearly shown in Table 2, there is a conflict between the total travel time and the total buffer time index of the road network; pursuing extremely high reliability often requires sacrificing some travel efficiency. Traffic managers can choose the appropriate option from the 21 options in Table 2 based on their current policy preferences: if prioritizing traffic efficiency, they can choose the option with the shorter total travel time; if greater emphasis is placed on travel reliability, they can choose the option with the shorter total buffer time index; if a balance between the two is required, a compromise option can be chosen in the middle of the Pareto front.

[0084] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for optimizing the layout of dedicated CAV lanes based on travel time reliability, characterized in that, include: Construct an expanded road network that includes both conventional and virtual road sections; Based on the expanded road network, the traffic capacity of regular road sections and virtual road sections is calculated according to the vehicle following mode and the corresponding safe headway. A mixed traffic flow allocation model is established based on the aforementioned traffic capacity, and the balanced traffic flow distribution of the road network is obtained by solving the mixed traffic flow allocation model. Based on the balanced traffic distribution, a set of travel time samples for road segments and routes is obtained through Monte Carlo random sampling; The road network buffer time index is calculated based on the travel time sample set, and the buffer time index characterizes the reliability of road network travel time. Using the number of CAV dedicated lanes on the virtual road segment as the decision variable, and taking the minimization of the total travel time and the minimization of the total buffer time index of the road network as the optimization objectives, a CAV dedicated lane layout optimization model is established. The CAV dedicated lane layout optimization model is used to output a set of CAV dedicated lane layout optimization schemes.

2. The method for optimizing the layout of CAV dedicated lanes based on travel time reliability according to claim 1, characterized in that, Constructing the extended road network includes: For each segment of the initial road network, a virtual segment is added, and the initial segment is designated as a regular segment. The sum of the number of lanes in the regular segment and the virtual segment is equal to the total number of lanes in the corresponding segment of the initial road network. The regular segment allows connected autonomous vehicles and human-driven vehicles to travel together, while the virtual segment only allows connected autonomous vehicles to travel.

3. The method for optimizing the layout of CAV dedicated lanes based on travel time reliability according to claim 1, characterized in that, The vehicle following mode includes: On the regular road sections, there are five modes: human-driven vehicles following human-driven vehicles, human-driven vehicles following connected autonomous vehicles, connected autonomous vehicles following human-driven vehicles, connected autonomous vehicles following connected autonomous vehicles in the same platoon, and connected autonomous vehicles following connected autonomous vehicles in different platoons. The virtual road segment is divided into two modes: connected autonomous vehicles following connected autonomous vehicles in the same platoon, and connected autonomous vehicles following connected autonomous vehicles in different platoons.

4. The method for optimizing the layout of CAV dedicated lanes based on travel time reliability according to claim 1, characterized in that, The traffic capacity of the aforementioned regular road section is: ； in, To ensure the traffic capacity of regular road sections, This refers to the number of lanes on a regular road section. This represents the average headway on a typical road section. , , , , Different vehicle following modes for regular road sections , , , for , , , , Corresponding to different safe headway distances, To expand the collection of regular road sections in the road network; The traffic capacity of the virtual road segment is: ； in, For the traffic capacity of virtual road segments, This represents the average headway on the virtual road segment. , Different vehicle following modes for virtual road segments. , for , Corresponding to different safe headway distances, This refers to the number of lanes on a virtual road segment. To expand the set of virtual road segments in the road network.

5. The method for optimizing the layout of CAV dedicated lanes based on travel time reliability according to claim 1, characterized in that, Based on the aforementioned traffic capacity, a mixed traffic flow allocation model is established. Solving the mixed traffic flow allocation model yields the equilibrium traffic distribution of the road network, including: Based on the aforementioned capacity, a variational inequality problem is constructed, which includes a path cost function, a flow conservation constraint, and a variable nonnegativity constraint. The path cost function is formed by aggregating the travel times of road segments, and the travel times of road segments are calculated using the BPR function based on the flow and capacity of the road segments. The variational inequality problem is solved using an iterative algorithm based on path exchange until the convergence condition is met, thereby obtaining the equilibrium flow distribution.

6. The method for optimizing the layout of CAV dedicated lanes based on travel time reliability according to claim 1, characterized in that, Based on the aforementioned balanced traffic distribution, the travel time sample set for road segments and routes obtained through Monte Carlo random sampling includes: S1. Based on the balanced traffic distribution, generate a random permutation sequence of connected autonomous vehicles and human-driven vehicles for each road segment; S2. Identify the following pattern between vehicles in the random sequence and assign a corresponding safe headway. Calculate the average headway of the road segment based on the safe headway of all vehicle pairs. S3. Calculate the road segment capacity based on the average headway of the road segment, and calculate the travel time of the road segment using the BPR function; S4. Calculate the route travel time based on the road segments included in the route and the travel time of each road segment. S5, Repeat S1-S4 Next, retrieve each Construct a set of route travel times from a set of different travel time samples.

7. The method for optimizing the layout of CAV dedicated lanes based on travel time reliability according to claim 1, characterized in that, The calculation of the road network buffer time index based on the aforementioned travel time sample set includes: Extract the 95th quantile from the sample set of path travel times as the planned travel time, and calculate the average value of the sample set of path travel times as the average travel time. The route buffer time index is obtained by dividing the difference between the planned travel time and the average travel time by the average travel time. The buffer time index of origin-destination pairs is obtained by taking a flow-weighted average of all path buffer time indices between origin-destination pairs in the road network. The total buffer time index of the road network is obtained by taking a flow-weighted average of the buffer time indexes for all origin and destination points.

8. The method for optimizing the layout of CAV dedicated lanes based on travel time reliability according to claim 1, characterized in that, The constraints of the CAV dedicated lane layout optimization model include: lane resource conservation constraints, traffic flow conservation constraints, user equilibrium constraints, and generalized travel cost constraints.

9. The method for optimizing the layout of CAV dedicated lanes based on travel time reliability according to claim 1, characterized in that, The CAV dedicated lane layout optimization model is used to output a set of CAV dedicated lane layout optimization schemes, including: solving the CAV dedicated lane layout optimization model using a fast elite multi-objective genetic algorithm; The CAV dedicated lane layout optimization model is solved using a fast elite multi-objective genetic algorithm, including: S1. Encode the virtual road segment layout scheme with integers and randomly generate an initial population; S2. The offspring population is generated through selection operations based on non-dominated sorting and crowding distance, simulated binary crossover operations, and polynomial mutation operations. S3. After merging the parent and offspring populations, perform rapid non-dominated sorting and select individuals to form a new generation population based on sorting level and crowding distance. S4. Iterate through S1-S3 until the maximum number of generations is reached, and output the solutions in the non-dominated levels as the set of CAV dedicated lane layout optimization solutions.

Images