A lane-level space-time resource allocation method for urban road network in an intelligent network environment

Abstract

The present application relates to the technical field of intelligent traffic control, and particularly relates to a kind of urban road network lane level space-time resource allocation methods under intelligent network connection environment, comprising: based on a plurality of signal intersections and a plurality of road segments to construct urban road network model;Determine the pollution bearing capacity and resource bearing capacity of a plurality of edges respectively, determine single side bearing capacity and urban road network bearing capacity based on the pollution bearing capacity and the resource bearing capacity;Step S3, based on the preset passable traffic flow in unit time, the single side bearing capacity and road network bearing capacity, lane level space resource pre-allocation is carried out to urban road network space resource;Step S4, construct hybrid traffic flow dataset, predict the short-time lane level passable state of urban road network based on the hybrid traffic flow dataset;Step S5, based on the result of the lane level space resource pre-allocation and the short-time lane level passable state, realize the optimal allocation of lane level space-time resource.The present application realizes urban road network lane level space-time resource allocation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent traffic control, and in particular to a lane-level space-time resource allocation method for urban road networks in an intelligent network environment. BACKGROUND

[0002] The improvement of the intelligent degree of urban traffic can improve the travel experience of the people and promote the sustainable development of the economy. Traffic safety is the top priority. With the acceleration of urbanization, the traffic flow of the city is increasing, especially during peak hours, and the traffic congestion phenomenon is becoming increasingly serious. How to tap the potential of collaborative perception of vehicles and traffic infrastructure, improve the efficiency of road network, and reduce traffic pollution has become an important problem to be solved in the intelligent development of urban traffic at the present stage.

[0003] The existing space-time resource allocation of urban road networks mainly involves intelligent network technology, urban road network modeling technology, urban road network short-term traffic flow prediction technology, and urban road network space-time resource allocation technology. Through reasonable space-time resource allocation, traffic congestion can be alleviated, thereby improving the efficiency of road traffic.

[0004] These technologies have the following shortcomings, including: (1) the existing space-time resource allocation of urban road networks is mainly aimed at specific urban road networks, and there is a lack of research on lane-level space-time resource allocation, and there is no method to refine the road network resources; (2) the coupling between the overall planning scheme of the road network and the local planning scheme of the vehicle in the existing space-time resource allocation scheme of the urban road network is poor, resulting in insufficient utilization of space-time resources and the inability to balance environmental protection and traffic efficiency. (3) The existing urban road network model is usually based on fixed and static data to complete road network modeling, without combining real-time dynamic lightweight data for real-time analysis; (4) the prediction effect of the short-term traffic flow prediction model is not ideal, the training time is long, and the accumulated error easily affects the accuracy of the short-term traffic flow prediction. SUMMARY

[0005] Therefore, the present application provides a lane-level space-time resource allocation method for urban road networks in an intelligent network environment to overcome the problems of not refining to the lane level in the existing space-time allocation of urban road networks, insufficient utilization of space-time resources, and low real-time performance and short-term traffic flow prediction accuracy.

[0006] To achieve the above purpose, on the one hand, the present application provides a lane-level space-time resource allocation method for urban road networks in an intelligent network environment, comprising:

[0007] Step S1, constructing an urban road network model based on a plurality of signalized intersections and a plurality of road segments, wherein the plurality of signalized intersections are a plurality of road network nodes, and the plurality of road segments are a plurality of edges;

[0008] Step S2, based on the urban road network model, respectively determine the pollution carrying capacity and resource carrying capacity of several edges, based on the pollution carrying capacity and the resource carrying capacity to determine the single edge carrying capacity and urban road network carrying capacity of several edges;

[0009] Step S3, based on the preset passable traffic flow per unit time, the single edge carrying capacity and road network carrying capacity, lane level space resource pre-allocation is performed on urban road network space resources;

[0010] Step S4, obtain real-time self-lane flow data, historical self-lane flow data, real-time adjacent lane flow data, historical adjacent lane flow data of several vehicles in the preset passable traffic flow, and combine road network environment event data to construct a hybrid traffic flow data set, and predict the short-time lane level passable state of the urban road network based on the hybrid traffic flow data set;

[0011] Step S5, based on the results of the lane level space resource pre-allocation and the short-time lane level passable state, the optimal allocation of lane level space-time resources is realized.

[0012] Further, in the step S1, comprising:

[0013] Step S11, determine the vehicle travel time of several vehicles between any two road network nodes;

[0014] Step S12, based on the vehicle travel time, predict the average vehicle travel time between any two road network nodes;

[0015] Wherein, the average vehicle travel time is the time cost of the vehicle passing through any two road network nodes.

[0016] Further, in the step S3, comprising:

[0017] Step S31, based on a preset proportion coefficient and a preset allocation number, determine the total amount of urban road network that can be allocated per unit time, and based on the preset allocation number, determine each OD data;

[0018] Step S32, based on the preset allocation number, iteratively order the each OD data, compare the iteration OD data with the urban road network carrying capacity per unit time, and according to the comparison result, adjust the total amount of urban road network that can be allocated and pre-allocate the urban road network space resources.

[0019] Further, in step S32, at each iteration, comprising:

[0020] Determine the current position of several vehicles, and based on the current position of several vehicles and the maximum transfer probability, determine the target road network node of several vehicles;

[0021] determining a single-edge carrying capacity of the several vehicles from the current position to a target road network node;

[0022] comparing the single-edge carrying capacity with the single-OD data, and determining a target road network node type according to a comparison result;

[0023] adjusting the target road network node of the several vehicles according to the target road network node type.

[0024] Further, in step S4, the adjacent lane of the several vehicles is determined, including:

[0025] determining a relevant lane set of the several vehicles based on a real-time driving section of the several vehicles;

[0026] calculating a correlation coefficient of a lane in the relevant lane set and a lane of the several vehicles, respectively;

[0027] comparing the correlation coefficient with a preset correlation threshold, and determining the adjacent lane of the several vehicles according to a comparison result.

[0028] Further, in step S4, a short-time lane-level passable state is predicted based on the mixed traffic flow data set, including:

[0029] predicting a short-time lane-level traffic flow based on a preset time period;

[0030] determining a lane-level passable state of a single road network node based on the short-time lane-level traffic flow and four directions of east, west, south and north;

[0031] determining a short-time lane-level passable state based on the lane-level passable state.

[0032] Further, a node weight matrix of the single road network node is determined, including:

[0033] determining a traffic cost of the single road network node based on an average vehicle speed of an upstream lane and an average vehicle speed of a downstream lane of the single road network node;

[0034] determining a signal control cost of the single road network node based on a traffic state and a traffic time cost of the single road network node;

[0035] determining the node weight matrix of the single node based on the traffic cost and the signal control cost;

[0036] The traffic state includes a red light state and a green light state.

[0037] Further, in step S5, a lane-level space-time resource optimization intermediate state matrix is determined based on the lane-level space resource pre-allocation result and a time-varying OD vector of the several vehicles in the urban road network.

[0038] Further, the optimal passing target of the plurality of vehicles is determined based on the lane-level space-time resource optimization intermediate state matrix.

[0039] Further, the general rule constraint and the optimal passing constraint are taken as constraint conditions of the optimal passing target.

[0040] Compared with the prior art, the beneficial effects of the present application are that by constructing a refined urban road network model, the pollution carrying capacity and the resource carrying capacity are comprehensively considered to evaluate the urban road network carrying capacity, and then the lane-level space resource is pre-allocated. At the same time, the mixed traffic flow data set is constructed by combining the real-time self-lane flow data, the historical self-lane flow data, the real-time adjacent lane flow data, the historical adjacent lane flow data and the road network environment event data to provide data support for accurately predicting the short-time lane-level passable state, finally optimizing the lane-level space-time resource allocation, effectively optimizing the urban road traffic operation mode, reducing pollution, utilizing the advantages of multi-source fusion technology, combining the lightweight traffic data set to provide microscopic allocation and optimization scheme for the urban road network, and improving the road network passing efficiency under the autonomous traffic system and the interaction ability between different autonomous levels.

[0041] Further, the present application determines the total amount of urban road network that can be allocated in a unit of time based on a preset proportion coefficient and a preset allocation number, and determines each OD data based on the preset allocation number, iterates the each OD data in order, adjusts the total amount of urban road network that can be allocated based on the comparison result of the urban road network carrying capacity and the iteration result, and pre-allocates the urban road network space resource, thereby further optimizing the urban road traffic operation mode, facilitating the combination with time characteristics in the subsequent, utilizing the advantages of multi-source fusion technology, combining the lightweight traffic data set to provide microscopic allocation and optimization scheme for the urban road network, and improving the road network passing efficiency under the autonomous traffic system and the interaction ability between different autonomous levels.

[0042] Further, the present application determines the correlation coefficient of the lane in the relevant lane set that the plurality of vehicles drive in real time based on the mixed traffic flow data set from the relationship between the lanes, determines the adjacent lane of the plurality of vehicles through the comparison result of the correlation coefficient and a preset correlation threshold, considers the time-varying lightweight data set, thereby further constructing the time-varying lane-level road network passability model through the short-time prediction of the passing flow of the lane, and further improving the road network passing efficiency under the autonomous traffic system and the interaction ability between different autonomous levels.

[0043] Further, the application combines the average vehicle speeds of the upstream lane and the downstream lane of a single road network node to evaluate the passing cost, and calculates the signal control cost according to the passing state and the passing time cost of the road network node, and then constructs the node weight matrix of the single road network node, which helps to comprehensively and accurately quantify the influence of the road network node in the traffic flow, identify the hotspot area of traffic congestion, thereby significantly reducing the passing delay, improving the road passing capacity and the overall traffic operation efficiency, and further improving the road network passing efficiency under the autonomous traffic system and the interaction ability between different autonomous level subjects.

[0044] Further, the application takes the general rule constraint and the optimal passing constraint as the constraint conditions of the optimal passing target, limits the road passing rationality and safety under the basic conditions of ensuring the vehicle passing in the urban road network, so as to realize the optimal passing target, and further obtain the optimal solution of the time-varying lane-level road network passability model, thereby further improving the road network passing efficiency under the autonomous traffic system and the interaction ability between different autonomous level subjects. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 The flowchart of the lane-level space-time resource allocation method of the urban road network under the intelligent network environment of the application;

[0046] Figure 2 The flowchart of the lane-level space-time resource allocation optimization algorithm of the embodiment of the application;

[0047] Figure 3 The schematic diagram of the mutual relationship between the traffic carrying capacity system elements of the embodiment of the application;

[0048] Figure 4 The topological node graph of the road network model G of the embodiment of the application;

[0049] Figure 5 The schematic diagram of the 2*3 road network structure of the embodiment of the application;

[0050] Figure 6 The method architecture diagram of the BI-LSTM short-term prediction method of the embodiment of the application. DETAILED DESCRIPTION

[0051] In order to make the objects and advantages of the application clearer, the application will be further described below in combination with embodiments; it should be understood that the specific embodiments described herein are only used to explain the application, and do not limit the protection scope of the application.

[0052] The preferred embodiments of the application will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the application, and are not intended to limit the protection scope of the application.

[0053] It should be noted that in the description of the present application, the terms of direction or positional relationship indicated by "upper", "lower", "left", "right", "inner", "outer" and the like are based on the direction or positional relationship shown in the drawings, which is only for the convenience of description, and does not indicate or imply that the device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the present application.

[0054] In addition, it should also be noted that in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixed connection, or detachable connection, or integral connection; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through intermediate medium, or the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0055] Please refer to Figure 1 and Figure 2 , Figure 1 is a flow chart of the method for allocating lane-level space-time resources of urban road network in the intelligent network environment of the present application, Figure 2 is a flow chart of the lane-level space-time resource allocation optimization algorithm of the embodiment of the present application. Specifically, the present application provides a method for allocating lane-level space-time resources of urban road network in the intelligent network environment, which comprises:

[0056] Step S1, constructing a city road network model based on a plurality of signalized intersections and a plurality of road segments, wherein the plurality of signalized intersections are a plurality of road network nodes, and the plurality of road segments are a plurality of edges;

[0057] Step S2, determining the pollution carrying capacity and the resource carrying capacity of the plurality of edges based on the city road network model, and determining the single-edge carrying capacity and the city road network carrying capacity based on the pollution carrying capacity and the resource carrying capacity;

[0058] Step S3, pre-allocating lane-level space resources of city road network space resources based on the preset passable traffic flow per unit time, the single-edge carrying capacity and the road network carrying capacity;

[0059] Step S4, acquiring real-time self-lane flow data, historical self-lane flow data, real-time adjacent lane flow data, historical adjacent lane flow data of a plurality of vehicles in the preset passable traffic flow, and combining road network environment event data to construct a hybrid traffic flow data set, and predicting the short-time lane-level passable state of the city road network based on the hybrid traffic flow data set;

[0060] Step S5, realizing the optimal allocation of lane-level space-time resources based on the results of the lane-level space resource pre-allocation and the short-time lane-level passable state.

[0061] It can be understood that the real two-way two-lane road environment can be topological into a directed graph G in a digital simulation environment, and a city road network model is established, and the formula is as follows:

[0062] G={N,E} (1)

[0063] Wherein, G is a set of city road network model; N is a set of several road network nodes in the city road network model; E is a set of several edges in the city road network model.

[0064] Please refer to Figure 3 It is the mutual relationship between the elements of the traffic carrying capacity system of the embodiment of the application, and it can be seen from the figure that the city road network traffic carrying capacity system contains three main parts: traffic operation system, traffic support system and traffic constraint system.

[0065] The city road network carrying capacity is determined based on the traffic constraint system, and the city road network carrying capacity can be divided into pollution carrying capacity and resource carrying capacity, and the calculation formula of pollution carrying capacity TEPCC and resource carrying capacity TERCC is as follows:

[0066] TEPCC=minAPC (2)

[0067] TERCC=min(alpha L LCC,alpha E ECC,alpha M MCC) (3)

[0068] Wherein, APC represents the atmospheric pollution carrying capacity, LCC represents the land resource carrying capacity, ECC represents the energy resource carrying capacity, MCC represents the mineral resource carrying capacity, alpha L , alpha E , alpha M Can be set according to different traffic scenarios.

[0069] The calculation formula of the city road network carrying capacity TECC is as follows:

[0070] TECC=min(TERCC,TEPCC) (4)

[0071] It can be understood that the pollution carrying capacity and resource carrying capacity of several edges, and the single edge carrying capacity can also be calculated based on the above formula.

[0072] The application realizes lane-level spatial resource pre-allocation by constructing a refined urban road network model, comprehensively considering pollution bearing capacity and resource bearing capacity to evaluate urban road network bearing capacity. Meanwhile, combined with vehicle real-time self-lane flow data, historical self-lane flow data, real-time adjacent lane flow data, historical adjacent lane flow data and road network environment event data, a hybrid traffic flow data set is constructed to provide data support for accurately predicting short-term lane-level passable state, finally optimizing lane-level space-time resource allocation, effectively optimizing urban road traffic operation mode and reducing pollution. By taking advantage of the multi-source fusion technology, a microscopic allocation and optimization scheme is provided for the urban road network by combining a lightweight traffic data set, so as to improve the road network passing efficiency under the autonomous traffic system and the interaction ability between different autonomous level subjects.

[0073] Specifically, in the step S1, the following steps are included:

[0074] Step S11, determining vehicle travel time of a plurality of vehicles between any two road network nodes;

[0075] Step S12, predicting vehicle average travel time between any two road network nodes based on the vehicle travel time;

[0076] The vehicle average travel time is the time cost of the vehicle passing between any two road network nodes.

[0077] In a specific embodiment, please refer to Figure 4 which is a topological node graph of the road network model G of the embodiment of the application. Specifically, based on a plurality of defined road network nodes and a plurality of edges, a projection coordinate system of the urban road network model G is constructed. The process of a vehicle from the lower left corner intersection to the upper right corner intersection in the urban road network model G is defined as the process from node N 01 to node N 43 , wherein the numerical values above and below the plurality of edges represent the lane-level predicted value of the vehicle average travel time from node N i to node N j , which is called time cost C i,j , and each lane has a separate time cost.

[0078] Specifically, in the step S3, the following steps are included:

[0079] Step S31, determining the total amount of urban road network that can be allocated in unit time based on a preset proportion coefficient and a preset allocation number, and determining each OD data based on the preset allocation number;

[0080] In step S32, the OD data is iterated in order based on the preset allocation times, the iteration OD data is compared with the urban road network carrying capacity in unit time, the urban road network allocatable total amount is adjusted according to the comparison result, and the urban road network space resource is pre-allocated.

[0081] It can be understood that the urban road network allocatable total amount in unit time is the total amount of vehicles passing through the urban road network in unit time.

[0082] In one specific embodiment, the preset proportion coefficient and the preset allocation times are determined based on the space-time consumption method, and the urban road network allocatable total amount in unit time is determined based on the preset proportion coefficient and the preset allocation times, and the related calculation formula is as follows:

[0083] q1+γq2+…+γ n-1 q n =Q(5)

[0084]

[0085] Wherein, the urban road network allocatable total amount Q in unit time is divided into several OD data {q1, q2, …, qn}, and the preset proportion coefficient is γ. n

[0086] According to the characteristics of the urban road network, a targeted maximum flow model is designed, the maximum capacity of the urban road network is calculated according to the multi-dimensional constraint condition, it is assumed that the traffic capacity of several road sections between nodes in the urban road network model is known, the maximum capacity of the urban road network and the minimum total cost of the urban road network are taken as the objective function, and the balanced flow of several road sections is less than the traffic capacity of several road sections as the constraint condition to solve the pre-allocation problem of the urban road network resource, and the expression of the model is as follows:

[0087]

[0088]

[0089] Wherein, X is the maximum traffic volume of the urban road network in unit time, x ij is the traffic volume between the starting point i and the ending point j, A is a set of several road sections, B a is the carrying capacity of the road section a, expressed as pcu / h, t a (x) is the impedance coefficient of the road section, h kij is the traffic volume of the road section k between the starting point i and the ending point j, is the use coefficient of the road section k between the starting point i and the ending point j, is the balanced flow of the road section a. Wherein, the road section k is the road section between the starting point i and the ending point j in the microscopic angle, and the road section a is the road section in the set A of several road sections in the macroscopic angle. ​

[0090] Please refer to Figure 5 as shown, Figure 5 Figure 2 is a schematic diagram of a 2x3 road network structure of an embodiment of the present application. In this city road network model, there are 6 road network nodes and 6 edges, and there are 2 routes that can be taken. Based on this, the total amount of city road network space resources that can be allocated is divided into 2 parts. The calculation formula of the total amount of city road network space resources that can be allocated is as follows:

[0091]

[0092] Among them, is the sum of the first m-1 parts of traffic, indicating that there are m road network nodes in the first route, and there are m-1 edges between the m road network nodes, and the m-1 edges correspond to m-1 parts of traffic. is the sum of the first t-1 parts of traffic, indicating that there are t road network nodes in the second route, and there are t-1 edges between the t road network nodes, and the t-1 edges correspond to t-1 parts of traffic. If the first loading traffic causes the city road network carrying capacity to be saturated, then the total number of routes is reduced to n (the total number of passable routes), and then the second loading amount of the city road network is calculated. Reduce to n, calculate the second loading amount of the road network.

[0093] It can be understood that when the routes selected by the vehicles are different, the number of road network nodes passed is different, and the total amount of city road network space resources that can be allocated is also different.

[0094] It can be understood that the iteration of each part of the OD data is in order, and the iteration of the OD data is compared with the city road network carrying capacity per unit time. If the iteration result is less than the city road network carrying capacity after iterating all the OD data, it means that there is still space for allocation of city road network space resources, so the total amount of city road network that can be allocated is increased, and Q is determined again.

[0095] In the implementation, the improved ant colony algorithm is used to pre-allocate lane-level space resources for city road network space resources. The lane-level space resource pre-allocation method can be adjusted according to the actual situation, which will not be described here.

[0096] The present application determines the total amount of city road network that can be allocated per unit time based on a preset proportion coefficient and a preset allocation number, and determines each part of the OD data based on the preset allocation number. The total amount of city road network that can be allocated is adjusted based on the comparison result of the city road network carrying capacity and the iteration result, and the city road network space resources are pre-allocated, so as to further optimize the city road traffic operation mode. Facilitate subsequent combination with time characteristics, take advantage of the advantages of multi-source fusion technology, combine lightweight traffic data set to provide microscopic allocation and optimization scheme for city road network, and improve the road network passing efficiency under the autonomous traffic system and the interaction ability between different autonomous level subjects.

[0097] Specifically, in step S32, at each iteration, the following is included:

[0098] determining a current position of each vehicle, determining a target road network node of each vehicle based on the current position of each vehicle and the maximum transition probability of each vehicle;

[0099] determining a single-lane carrying capacity of each vehicle from the current position to the target road network node;

[0100] comparing the single-OD data with the single-lane carrying capacity, and determining a target road network node type according to a comparison result;

[0101] adjusting the target road network node of each vehicle according to the target road network node type.

[0102] In one specific embodiment, when lane-level spatial resource pre-allocation is performed on the spatial resources of the urban road network, the set of visits is defined as the road network nodes to be visited by the vehicles, the set of visited is defined as the road network nodes that have been visited by the vehicles, and the number of vehicles is n.

[0103] The maximum transition probability of each vehicle is determined based on the current position of each vehicle to determine the target road network node of each vehicle.

[0104] The transition probability of the kth individual when the road network node i goes to the target road network node j The calculation formula is as follows:

[0105]

[0106] wherein, represents the transition probability of the kth vehicle when the road network node i goes to the target road network node j, τ i,j represents the pheromone content of the road network node i going to the target road network node j (the pheromone content is represented as the number of times of passing through by a single vehicle). η i,j represents the visibility, which is the expected degree of the road network node i going to the target road network node j. α and β represent the importance parameters of factors.

[0107] Based on the above transition probability calculation, the maximum transition probability of each vehicle can be obtained, and the next target road network node of each vehicle can be selected based on the maximum transition probability. After the target road network node is determined, the single-OD data is compared with the single-lane carrying capacity, and the target road network node type is determined according to the comparison result. If the single-OD data is greater than the single-lane carrying capacity, it is determined that the target road network node type is visited, and is put into the set of visited. Then, the next optimal target road network node is searched.

[0108] In the new round of iteration process, the pheromone content is recalculated, and the calculation formula is as follows:

[0109]

[0110]

[0111] Wherein, represents the pheromone content of the road network node i to the target road network node j in the new round of iteration process, represents the pheromone content of the node i to the target node j in the last round of iteration process. P represents the volatility degree of pheromone, which is mapped to the weight of reducing the long road section in the vehicle space resource allocation. represents the total increment of pheromone of each road section after the vehicle I traverses the path once, O m represents the total consumption of the vehicle I on the road section, also known as the passing cost.

[0112] Specifically, in step S4, the adjacent lanes of the vehicle are determined, including:

[0113] Determine the relevant lane set of the vehicle based on the real-time driving road section of the vehicle;

[0114] Calculate the correlation coefficient of the lanes in the relevant lane set and the lane of the vehicle itself, respectively;

[0115] Compare the correlation coefficient with the preset correlation threshold, and determine the adjacent lane of the vehicle according to the comparison result.

[0116] It can be understood that the real-time driving road section of the vehicle is set as "upstream road section", and the relevant lane set is the other lanes in the same "upstream road section" as the lane of the vehicle itself when driving in real time. The correlation between the lanes in the relevant lane set and the lane of the vehicle itself is determined by calculating the correlation coefficient of the lanes in the relevant lane set and the lane of the vehicle itself.

[0117] The calculation formula of the correlation coefficient is as follows:

[0118]

[0119] Wherein, σ L1 and σ L2 respectively represent the standard deviation of the traffic flow of roads L1, L2. cov(L1, L2) is the covariance between L1 and L2, and n is the number of historical data, and respectively represent the average traffic flow.

[0120] It can be understood that if the correlation coefficient is greater than a preset correlation threshold, the lanes in the relevant lane set are determined as adjacent lanes of the vehicles.

[0121] In implementation, the preset correlation threshold has a value range of 0.4-0.6, preferably, the preset correlation threshold is 0.5, and the value range and the preferred value of the preset correlation threshold can be adjusted according to actual conditions, which will not be described here.

[0122] The present application determines the correlation coefficient of the lanes in the relevant lane set of the vehicles in real time based on the mixed traffic flow data set from the relationship between the lanes, determines the adjacent lanes of the vehicles through the comparison result of the correlation coefficient and the preset correlation threshold, considers the time-varying lightweight data set, and further constructs the time-varying lane-level road network accessibility model through the short-time predicted lane-level traffic flow, thereby further improving the road network traffic efficiency under the autonomous traffic system and the interaction ability between different autonomous level subjects.

[0123] Specifically, in step S4, the short-time lane-level passable state is predicted based on the mixed traffic flow data set, including:

[0124] The short-time lane-level traffic flow is predicted based on a preset time period;

[0125] The lane-level passable state of a single road network node is determined based on the short-time lane-level traffic flow and the four directions of east, west, south and north;

[0126] The short-time lane-level passable state is determined based on the lane-level passable state.

[0127] It can be understood that the present application starts from the relationship between the lanes, considers the time-varying lightweight data set, adopts the real-time lane flow data, the historical lane flow data, the real-time adjacent lane flow data, the historical adjacent lane flow data of the vehicles and combines the road network environment event data to construct the mixed traffic flow data set, and proposes a BI-LSTM short-time prediction method to predict the short-time lane-level traffic flow in a preset time period.

[0128] Please refer to Figure 6 shown, Figure 6 which is a method architecture diagram of the BI-LSTM short-time prediction method of the embodiment of the present application. After the mixed traffic flow data set is constructed, it is input to two LSTM layers in a forward order and a reverse order respectively for feature extraction, and a dense layer (Dense Layer) is further added to improve the prediction accuracy.

[0129] In implementation, the preset time period is 5 min, 10 min or 20 min, and the value range of the preset time period is not specifically limited as long as the short-time lane-level traffic flow prediction can be realized, which will not be repeated here.

[0130] The results of the Bi-LSTM prediction of the short-time lane-level traffic flow of 5 min, 10 min and 20 min are weighted and superimposed, the prediction results of short time are adjusted to a higher priority level, and the prediction results of long time are adjusted to a lower priority level, to generate the passable state matrix of the road network.

[0131]

[0132]

[0133]

[0134] wherein Y represents the digitized lane-level passable state (lane passable traffic volume), e, s, w and n represent the east, south, west and north directions respectively, input and output represent the import lane and export lane respectively, and the numbers 1 and 2 represent the first lane and the second lane in the direction from south to north or from west to east. Formula (15) represents the passable state of the first node lane, and the matrix represented by formula (15) is expanded by row and combined with the rest of the nodes of the road network to obtain the road network lane-level passable state matrix as shown in formula (16).

[0135] Specifically, the node weight matrix of the single road network node is determined, including:

[0136] The passage cost of the single road network node is determined based on the upstream lane average speed and the downstream lane average speed of the single road network node;

[0137] The signal control cost of the single road network node is determined based on the passage state and the passage time cost of the single road network node;

[0138] The node weight matrix of the single node is determined based on the passage cost and the signal control cost.

[0139] The passage state includes the red light state and the green light state.

[0140] In a specific embodiment, the upstream and downstream speeds of the i-th lane of the intersection of the city road network at time t are defined as V t i The formula is as follows:

[0141]

[0142] wherein v i,j up(t) represents the average speed of the lane j upstream of the intersection i at time t, v i,j down (t) represents the average speed of the lane j downstream of the intersection i at time t, N lane represents the total number of import lanes.

[0143] Define the travel cost of the lane N lane upstream and downstream of the node i (representing the average speed state of the lane level), assuming that the vehicle travels from upstream to downstream of the node i, the travel cost formula is as follows:

[0144]

[0145] Assuming that at time t, the vehicle travels at the current speed to the stop line in front of the node (signalized intersection) of the road network at a constant speed and takes Δt. The yellow light time is calculated as the green light time, and has the ability to pass through the current signalized intersection. The travel state of the vehicle at the signalized intersection is simplified as a green light state (passable state) and a red light state (unpassable state). When the state of the signalized intersection in front of the vehicle at time t0+Δt is unpassable, the fixed travel cost is defined as C red . The remaining time of the red light at time t0is t red , and the intersection travel time cost is When the state of the intersection in front of the vehicle at time t0+Δt is passable, the current fixed travel cost is C green , the remaining time of the green light at time t0is t green , the total time of the current phase green light is G light , and the intersection travel time cost is The signalized cost of the vehicle at the signalized intersection (the signalized cost is the travel time cost) can be obtained, and the formula is as follows:

[0146]

[0147] Therefore, considering the travel cost and the signalized cost, the node weight matrix of the single road network node is determined, and the formula is as follows:

[0148]

[0149] wherein w through , w light respectively represent the weight coefficients of the travel cost and the signalized cost, which are determined according to the actual vehicle road conditions, and will not be described herein.

[0150] The application combines the average vehicle speeds of the upstream lane and the downstream lane of a single road network node to evaluate the passing cost, calculates the signal control cost according to the passing state and passing time cost of the road network node, and further constructs the node weight matrix of the single road network node, which helps to comprehensively and accurately quantify the influence of the road network node in the traffic flow, identifies the hotspot area of traffic congestion, thereby significantly reduces the passing delay, improves the road passing capacity and the overall traffic operation efficiency, and further improves the road network passing efficiency under the autonomous traffic system and the interaction ability between different autonomous level subjects.

[0151] Specifically, in step S5, the lane-level space-time resource optimization intermediate state matrix is determined based on the lane-level space resource pre-allocation result and the time-varying OD vector of the vehicles in the urban road network.

[0152] It can be understood that, in combination with the lane-level space resource allocation result in space, the lane-level space resource allocation strategy for the urban road network can be endowed with time characteristics, thereby determining the lane-level space-time resource optimization intermediate state matrix.

[0153] In one specific embodiment,

[0154] The time-varying OD vector of the vehicles in the urban road network is calculated according to the following formula:

[0155]

[0156] The path corresponding to each pair of OD information is:

[0157]

[0158] The lane-level space-time resource optimization intermediate state matrix (including the corresponding lane and the amount of resource to be allocated in the path selected by the vehicle) is as follows:

[0159]

[0160] Wherein, S OD represents the time-varying OD vector of the vehicle cluster in the urban road network, I n represents the nth vehicle, represents the path set corresponding to the OD information of the vehicle, and different time instants are endowed with different meanings, as shown in formula (25). represents the second north-south lane of the west-east direction of the nth intersection, M resources represents the lane-level space-time resource optimization intermediate state, represents the traffic volume corresponding to the second north-south lane of the west-east direction of the nth intersection at time t, represents the traffic volume corresponding to the second west-east lane of the north-south direction of the nth intersection at time t.

[0161] Specifically, the optimal passing target of the vehicles is determined based on the lane-level space-time resource optimization intermediate state matrix.

[0162] It can be understood that after the lane space-time resources of the urban road network are allocated to each vehicle, the optimal passing target of the vehicles is determined, the optimal passing target being the least congested path selected by the vehicles and the minimum total queue length of the entire urban road network, and the formula being as follows:

[0163]

[0164] wherein ω crowd represents the weight of the least congested path selection sub-item, ω queue represents the weight of the minimum total queue length sub-item, and ω crowd <ω queue . represents the length of the congested road section at time t, represents the average speed corresponding to the current traffic flow of the road, represents the traffic flow loaded on the road, h s represents the headway, M represents the total number of vehicles in the road network, and N represents the number of congested road sections in the vehicle OD information.

[0165] Specifically, the general rule constraint and the optimal passing constraint are taken as the constraint conditions of the optimal passing target.

[0166] It can be understood that the general rule constraint refers to the basic evolution rule of the urban road network traffic flow and the basic rule of the passable modeling, and the basic conditions that need to be ensured by the vehicles in the urban road network passing.

[0167] In a specific embodiment, the general rule constraint is as follows:

[0168] The first item indicates that each defined lane in the urban road network loads the lane-level traffic flow data, the second item indicates that each existing OD data must be loaded into the urban road network in the new round of calculation, the third item indicates that the traffic flow data loaded on each lane is not less than zero, the fourth item indicates that the necessary condition for the current lane traffic flow to pass through the current signalized intersection is that the average speed is less than the remaining green time, and the fifth item defines the independent variables in the first four items.

[0169]

[0170] wherein N assign represents the total number of lanes that can be allocated in the current urban road network, N node represents the total number of signalized intersections in the current urban road network, N direction represents the total number of driving directions existing in the urban road network, and Nlane represents the total number of lanes in the fixed driving direction of the current signalized intersection, represents the road allocated at time t, S close represents the OD information of the model being tuned, represents the traffic flow loaded on the kth lane in the jth driving direction of the ith intersection at time t, represents the length of the kth lane in the jth driving direction of the ith intersection, represents the current green light remaining time of the kth lane in the jth driving direction of the ith intersection.

[0171] It can be understood that the optimal traffic rule serves the optimal traffic target while limiting the rationality and safety of road traffic, and the optimal traffic rule is as follows:

[0172] The first term represents that the sum of the congestion degree of the new round of path allocation of the vehicle is less than the original path (which is a necessary condition for switching paths), the second term represents that the lane with higher congestion degree will be allocated less traffic flow, the third term represents the average speed limit of the entrance lane / exit lane of the signalized intersection, the fourth term represents the change rate limit of the traffic flow allocated to each lane by the space-time resource optimization method, the fifth term limits the change frequency of the optimal route update of the vehicle, and the sixth term represents that the selected lane in the model exists in the OD information of the vehicle.

[0173]

[0174] wherein, and respectively represent the congestion degree of the allocated road at time t and time t+1, and respectively represent the resource amount allocated to the nth lane and the n+1th lane at time t, represents the maximum average traffic flow speed constraint, ε q represents the change rate of the allocated resource amount, ε t represents a fixed time interval, represents the OD information of the vehicle at time t+1.

[0175] In a specific embodiment, the optimal solution (the time-labeled road network space resource allocation matrix) is obtained by the Pontryagin maximum method, and is allocated to each vehicle. The allocation strategy takes a single lane as the traffic space, allocates the bearable capacity of each traffic space at each time, and the lane-level space-time resource optimization result (the time-labeled road network space resource allocation matrix) is as follows:

[0176]

[0177] wherein, M assignrepresent the lane level space-time resource optimization result, T cur represent the current time, T arr represent the time to reach the next road network node, represent the time expected to be needed through the current road network node. Wherein, for each pair of OD information L OD corresponding to a number of vehicles in the city road network, a set of passable space with time label is allocated vehicles pass through the corresponding space node according to the time label at the specified time, that is, complete a space-time resource allocation, when the vehicles in the city road network complete all space-time resource allocation according to the space-time resource allocation matrix with time label, the allocation result is the optimal allocation result of the city road network vehicle driving path.

[0178] The application takes the general rule constraint and the optimal pass constraint as the constraint condition of the optimal pass target, limits the road pass rationality and safety under the basic condition of ensuring the vehicle pass in the city road network, so as to realize the optimal pass target, further find the optimal solution of the time-varying lane level road network passability model, and further improve the road network pass efficiency under the autonomous traffic system and the interaction ability between different autonomous level subjects.

[0179] So far, the technical scheme of the application has been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the application, and the technical scheme after the changes or replacements will fall within the protection scope of the application.

Claims

1. A lane-level space-time resource allocation method for urban road networks in an intelligent connected environment, characterized in that, The method comprises the following steps: Step S1, constructing a city road network model based on a plurality of signalized intersections and a plurality of road segments, wherein the plurality of signalized intersections are a plurality of road network nodes, and the plurality of road segments are a plurality of edges; Step S2, determining the pollution carrying capacity and the resource carrying capacity of the plurality of edges based on the city road network model, and determining the single-edge carrying capacity and the city road network carrying capacity based on the pollution carrying capacity and the resource carrying capacity; Step S3, performing lane-level spatial resource pre-allocation on the spatial resources of the city road network based on the preset passable vehicle flow per unit time, the single-edge carrying capacity and the city road network carrying capacity; In the step S3, the method comprises the following steps: Step S31, determining the total allocatable amount of the city road network per unit time based on a preset proportion coefficient and a preset allocation number, and determining each OD data based on the preset allocation number; Step S32, iteratively processing the OD data in sequence based on the preset allocation number, comparing the iterated OD data with the city road network carrying capacity per unit time, and adjusting the total allocatable amount of the city road network according to the comparison result and pre-allocating the spatial resources of the city road network; In the step S32, in each iteration, the method comprises the following steps: determining the current positions of a plurality of vehicles, determining the target road network nodes of the plurality of vehicles based on the current positions of the plurality of vehicles and the maximum transfer probability; determining the single-edge carrying capacity of the plurality of vehicles from the current positions to the target road network nodes; comparing the single OD data with the single-edge carrying capacity, and determining the target road network node type according to the comparison result; adjusting the target road network nodes of the plurality of vehicles according to the target road network node type; Step S4, obtaining real-time self-lane flow data, historical self-lane flow data, real-time adjacent-lane flow data, historical adjacent-lane flow data of the plurality of vehicles in the preset passable vehicle flow, and combining road network environmental event data to construct a hybrid traffic flow data set, and predicting the short-time lane-level passable state of the city road network based on the hybrid traffic flow data set; Step S5, realizing optimal allocation of lane-level space-time resources based on the results of the lane-level spatial resource pre-allocation and the short-time lane-level passable state.

2. The method of claim 1, wherein, In the step S1, the method comprises the following steps: Step S11, determining the vehicle travel time between any two road network nodes; Step S12, predicting the average vehicle travel time between any two road network nodes based on the vehicle travel time; The average vehicle travel time is the time cost of the vehicle passing between any two road network nodes. 3.The method of claim 1, wherein, In step S4, determining the adjacent lane of the plurality of vehicles comprises the following steps: determining the relevant lane set of the plurality of vehicles based on the real-time travel road segments of the plurality of vehicles; calculating the correlation coefficients of the lanes in the relevant lane set and the self-lane of the plurality of vehicles respectively; comparing the correlation coefficients with a preset correlation threshold, and determining the adjacent lane of the plurality of vehicles according to the comparison result.

4. The method of claim 1, wherein, In step S4, predicting the short-time lane-level passable state based on the hybrid traffic flow data set comprises the following steps: predicting the short-time lane-level passable flow based on a preset time period; determining the lane-level passable state of a single road network node based on the short-time lane-level passable flow and the four directions of east, west, south and north; Determine a short-time lane-level passable state based on the lane-level passable state.

5. The method of claim 4, wherein, Determine a node time-varying weight matrix of the single road network node, including: Determine a passable cost of the single road network node based on an upstream lane average speed and a downstream lane average speed of the single road network node; Determine a signal control cost of the single road network node based on a passable state and a passable time cost of the single road network node; Determine a node time-varying weight matrix of the single road network node based on the passable cost and the signal control cost; The passable state includes a red light state and a green light state.

6. The method of claim 1, wherein, In step S5, include: based on the lane-level space resource pre-allocation result and the time-varying OD vector of the plurality of vehicles in the urban road network to determine the lane-level space-time resource optimization intermediate state matrix.

7. The method of claim 6, wherein, Including: Determine the optimal passable target of the plurality of vehicles based on the lane-level space-time resource optimization intermediate state matrix.

8. The method of claim 7, wherein, Including, based on the general rule constraint and the optimal passable constraint as the constraint condition of the optimal passable target.

Images