Multi-intersection IVCPS cooperative control test case generalization generation method

Abstract

The invention belongs to the field of intelligent traffic, and relates to a multi-intersection IVCPS cooperative control test case generalization generation method, which comprises the following steps: S1, constructing a mixed traffic flow fundamental diagram model, and deducing the relationship between traffic flow density and flow under different permeability by combining IDM and CACC models; s2, expanding an IVCPS scene body, and designing a multi-intersection road network scene; s3, adopting an upper-layer and lower-layer distributed cooperative control architecture, wherein the upper layer dynamically adjusts signal timing based on traffic pressure information, and the lower layer provides an optimal path and driving speed suggestion for the CAV; s4, constructing a scene parameter extraction module, and extracting a control algorithm, a vehicle model ratio, traffic flow and CAV permeability; S5, constructing a key calculation module, and calculating a parameter key index; and S6, constructing a test case generation module, and generating a combined test case set by adopting an AETG-S algorithm. According to the invention, the traffic efficiency of a multi-intersection road network can be effectively improved, traffic delay and energy consumption are reduced, and a scientific basis is provided for testing and optimization of an intelligent traffic system.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation technology, and in particular relates to a method for generalizing and generating test cases for collaborative control of multi-intersection IVCPS. Background Art

[0002] With the acceleration of urbanization and the continued growth of car ownership, urban traffic congestion is becoming increasingly serious. Traditional traffic signal control methods are no longer able to meet the needs of complex traffic environments. Intelligent vehicle cyber-physical systems (IVCPS), as a key component of intelligent transportation systems, optimize the allocation of traffic resources through vehicle-road collaborative technology, becoming an effective means to alleviate traffic congestion and improve traffic efficiency. However, existing technologies still have the following shortcomings in multi-intersection collaborative control:

[0003] 1. Limitations of Traffic Flow Modeling: Traditional traffic flow models are often designed for a single vehicle type and struggle to accurately describe the characteristics of mixed traffic flows consisting of human-driven vehicles (HVs) and connected vehicles (CAVs). Existing models lack quantitative analysis of the dynamic impact of penetration, resulting in control strategies that are insufficiently adaptable to mixed CAV and HV scenarios.

[0004] 2. Insufficient coordination in the control architecture: Existing signal control methods (such as fixed timing and sensor-based control) typically optimize individual intersections independently and lack network-level coordination. The separation of vehicle routing and signal control further reduces overall system efficiency, potentially leading to localized congestion and resource waste.

[0005] 3. Limited coverage of test verification: Testing of multi-intersection IVCPS mostly relies on full-combination scenarios. The use cases are large-scale and inefficient, making it difficult to cover the combined impact of key parameters such as penetration rate and vehicle model ratio, which restricts the optimization and verification of control algorithms.

[0006] 4. The conflict between real-time response and global optimization: While edge computing devices can achieve real-time control, their global optimization capabilities are limited. Meanwhile, global optimization in the cloud often struggles to meet real-time requirements due to communication delays. Existing technologies have yet to effectively balance the needs of real-time response and global optimization. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide a generalized test case generation method for multi-intersection IVCPS collaborative control, which solves the above technical bottlenecks and achieves a comprehensive improvement in traffic efficiency and safety by constructing a mixed traffic flow model, designing a hierarchical collaborative architecture and an intelligent testing system.

[0008] In order to achieve the above object, the present invention provides the following technical solutions:

[0009] A generalized test case generation method for multi-intersection IVCPS collaborative control includes the following steps:

[0010] S1. Construct a basic graph model of mixed traffic flow, combine the IDM and CACC models, and derive the relationship between traffic flow density and flow under different penetration rates;

[0011] S2. Based on the basic graph model of mixed traffic flow, expand the IVCPS scenario ontology and design a multi-intersection road network scenario;

[0012] S3. A two-layer distributed collaborative control architecture is used: the upper layer dynamically adjusts signal timing based on traffic pressure information, while the lower layer provides optimal route and speed recommendations for CAVs.

[0013] S4. Build a scenario parameter extraction module to extract control algorithms, vehicle type ratios, traffic flow, and CAV penetration rates:

[0014] S5. Build a key calculation module to calculate the key indicators of parameters;

[0015] S6. Build a test case generation module and use the AETG-S algorithm to generate a combination test case set.

[0016] Furthermore, step S1 includes the following sub-steps:

[0017] S1.1 Construct a basic graph model of mixed traffic flow and calculate the average headway and flow-density relationship of mixed traffic flow;

[0018] The traffic flow density and flow rate of HV and CAV are:

[0019]

[0020] Where v represents the average speed of HV and CAV; k H represents the traffic density of HV; q H represents the flow rate of HV; v0 represents the expected speed of HV and CAV vehicles; s0 is the minimum safe distance between HV and CAV when they are stationary; T H represents the safe headway of HV; L represents the vehicle length; k C represents the traffic flow density of CAV; q C represents the flow rate of CAV; T C It represents the safe headway of CAV;

[0021] Assuming that the mixed traffic flow consists only of human-driven vehicles HV and intelligent connected vehicles CAV, the average headway h in the equilibrium state of the mixed traffic flow is derived: M , h M The calculation expression is:

[0022] hM =p·h C +(1-p)·h H

[0023] Where p represents the penetration rate of intelligent connected vehicles in the total vehicle population; 1-p represents the penetration rate of human-driven vehicles in the total vehicle population; h H Indicates the distance between HV vehicles; h C Indicates the distance between CAV heads;

[0024] The hybrid traffic flow basic graph model is constructed by the isomorphic traffic flow basic graph model. The hybrid traffic flow basic graph model is:

[0025]

[0026] Where k M represents the traffic density of mixed traffic flow; q M represents the flow rate of mixed traffic flow;

[0027] S1.2 Dynamically adjust the headway ratio between CAV and HV based on the penetration rate p;

[0028] The IDM model is used to describe the following behavior of human-driven vehicles, and the ACC or CACC model is used to describe the following behavior of intelligent connected vehicles. When the traffic flow is balanced, different types of vehicles maintain the same driving speed v, but the following distances maintained by human-driven vehicles and intelligent connected vehicles are different. The equilibrium speed satisfies:

[0029]

[0030] Substituting the above formula into the core formula of IDM and CACC, we can get the HV headway h H Distance from CAV front h C The calculation expression is:

[0031]

[0032] h C =L+s0+T C v.

[0033] Furthermore, in step S2, the design of the multi-intersection road network scenario includes the following sub-steps:

[0034] I. Clarify design goals and scale conversion;

[0035] Clarify the expansion goal from a single intersection to a macro road network, and couple microscopic behaviors to macroscopic characteristics through the traffic flow model IDM / CACC;

[0036] II. Expanding IVCPS scenario ontology;

[0037] Based on the IVCPS structured model, a new dynamic object layer, static road layer, and coordination control layer are added. Through layered expansion, unified modeling of microscopic details and macroscopic characteristics is achieved;

[0038] III. Construct road network model;

[0039] The road network structure is represented by a directed graph G = (J, L), where the intersection set JJ and the road set LL containing entrances and exits constitute nodes and edges. Vehicle path planning is based on the feasible path set Rj and relies on a two-layer architecture of edge cloud and regional cloud to achieve dynamic resource allocation and collaborative management.

[0040] IV. Design collaborative control mechanisms;

[0041] Signal coordination is achieved through sub-area division and phase difference optimization, dynamically adjusting green light duration. CAVs utilize V2X communication for route and speed guidance, while HVs rely on traditional sensors.

[0042] V. Verification and testing;

[0043] System performance is verified through combined test cases, with evaluation indicators including macro and micro. Test results are fed back to model optimization to form a closed-loop design process.

[0044] Furthermore, in step S2, for human driving, an intelligent driver model (IDM) is selected to simulate the driver's acceleration and deceleration behaviors during the car-following process to reproduce various phenomena in traffic flow. The core formula of the IDM is:

[0045]

[0046] Where a n is the acceleration of vehicle n; a max is the maximum acceleration of the vehicle; v n is the current speed of the vehicle; v0 is the desired speed of the vehicle; s n is the actual distance between the vehicle and the preceding vehicle, s0 is the minimum safe distance when stationary, T H is the safe headway; b is the comfortable deceleration; Δv n is the speed difference between the vehicle and the preceding vehicle; L is the vehicle body length.

[0047] Furthermore, in step S2, a cooperative adaptive cruise control (CACC) model is selected for the smart car, and the core formula of CACC is:

[0048]

[0049] Where a i (t) is the acceleration of the vehicle at time t; a i-1(t) is the acceleration of the preceding vehicle; e i (t) is the following error; k is the rate of change of the following error; p and k d is the proportional-derivative control gain; p i (t) is the position of the vehicle at time t; p i-1 (t) is the position of the preceding vehicle at time t; L is the vehicle length; s0 is the minimum safe distance when the vehicle is stationary; T C is the preset headway; v i (t) is the speed of the vehicle at time t.

[0050] Furthermore, step S3 includes the following sub-steps:

[0051] S3.1 Estimate the length of the queue at the intersection;

[0052] The estimated queue length at a signalized intersection is defined as follows:

[0053] q i,o (t+1)=[q i,o (t)-α i,o (t)·(s i,o (t)-l i,o (t))+d i,o (t)+a i,o (t)] +

[0054] Where q i,o (t+1) is the road L in a certain direction of the intersection i , and L o →L i is the estimated value of the queue length at the next moment in the driving direction; q i,o (t) is the current queue length; α i,o (t) is L o →L i Release coefficient at time t in the direction; s i,o (t) is the road L i Saturation flow; l i,o (t) is the flow loss; d i,o (t) planning the vehicle flow that will pass through this road section and arrive soon for other routes in the road network; a i,o (t) is the road L that can be directly reached from outside the road network i Vehicle flow, set to 0 when external flow is not considered; [·] + For positive operations, when the result is negative, the result is recorded as 0, because the queue length cannot be negative;

[0055] S3.2 dynamically adjusts the phase green light duration;

[0056] Taking queue length as the metric in the traffic pressure formula, we get the core formula of traffic pressure:

[0057]

[0058] Where q j,i For intersection L i →L j Direction Road L j The queue length of road L j For road L i Downstream of; in phase I road L i There are multiple downstream roads, phaseI is the set of downstream roads; λ i,o For L i →L j Steering ratio;

[0059] Traffic lights are controlled using a cyclic control method based on traffic pressure. Based on a fixed control cycle and phase sequence, green light time is allocated according to the proportion of traffic pressure in all phases. The allocation formula is as follows:

[0060]

[0061] Where g i is the green light time of phase i, C is the signal period, p N is the number of signal phases, g min is the minimum green light time;

[0062] S3.3 uses a breadth-first search algorithm to optimize path selection;

[0063] The cost increment for each downward search step of the path is defined as follows:

[0064]

[0065] Where γ is the relative importance weight, which is used to balance the estimated queue length and road length; q L is the queue length at the current moment; l(L) is the road length;

[0066] For a path R, the total cost function is defined as follows:

[0067]

[0068] The goal of optimizing the search is as follows:

[0069]

[0070] Indicates the path from the root node J0 to the destination node J d Among the feasible paths, select the total cost The smallest path;

[0071] S3.4 Recommendations for coordinated speed guidance at multiple intersections;

[0072] Usually, a speed that is optimal in terms of driving comfort and traffic flow stability is selected, and safety margins are considered. The recommended speeds for each intersection are as follows:

[0073]

[0074] Where, d i is the distance to the i-th intersection ahead; t g,i It represents the remaining green light time (in seconds) of the current phase of the i-th intersection when the vehicle arrives at the i-th intersection; Δt represents the compensation reaction time.

[0075] Furthermore, the key indicators of the calculation parameters in step S5 include enhanced time to collision (ETTC) and post-entry time (PET);

[0076] The calculation expression of ETTC is:

[0077]

[0078] Where Ra is the relative distance between the two vehicles; Δv is the relative velocity; Δa is the relative acceleration;

[0079] The calculation expression of PET is:

[0080] PET=t f -t b

[0081] Where, t f The time when the vehicle that passes first leaves the conflict area; t b The time when the last vehicle enters the conflict area.

[0082] Furthermore, the specific content of step S6 is:

[0083] The control algorithm, vehicle type ratio, traffic flow, and CAV penetration rate are selected as scenario parameters. The relative criticality between the parameters is calculated to represent the relative criticality between the generated test cases. The relative criticality index between the values ​​of the four traffic flow elements is obtained by calculation, and the formula is as follows:

[0084]

[0085] Where, is the scene parameter; For importance; The exposure frequency reflects the probability of the scene concept attribute appearing in the actual road scene; is the scene parameter Is a specific value; Ω T is the set of coordinates of all concept nodes on the backtracking path T; w ij is the relative importance of the corresponding coordinate concept node; Indicates that the scene element value during the observation period is The total time of the scene element in the observation space is The total frequency of f total Indicates the total time of the observation period or the total frequency of the observation space;

[0086] With coverage strength t=2, the AETG-S algorithm is used to generate a set of combined test cases.

[0087] Beneficial effects:

[0088] 1. The integrated thinking solution of scale-decoupling and cross-scale calling proposed in this invention can partially solve the problems of the association of multiple-scale coupled systems in complex scenarios, the complex interoperability relationships, and the inability to accurately analyze interdependencies at a single scale. It enhances the reliability of beyond-visual-range predictive perception of multiple intelligent transportation participating entity systems, and provides richer information and a more robust foundation for planning and control in dynamic environments.

[0089] 2. Provide theoretical support and architectural guidance for data-driven mountain highway and tunnel traffic systems and key vehicle predictive cruise cloud control methods from the perspective of IVCPS.

[0090] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 This is a flow chart of a generalized test case generation method for multi-intersection IVCPS collaborative control according to the present invention;

[0092] Figure 2 This is the basic graph model of mixed traffic flow in this embodiment;

[0093] Figure 3 This is the basic graph model of mixed traffic flow under different permeabilities in this embodiment;

[0094] Figure 4 This is a partial ontology concept and attribute diagram under the large scale of IVCPS in this embodiment;

[0095] Figure 5 This is the IVCPS multi-intersection road network model of this embodiment;

[0096] Figure 6 This is the full-sensing control flow chart of this embodiment;

[0097] Figure 7 This is a schematic diagram of the sensing control parameters of this embodiment;

[0098] Figure 8 This is a schematic diagram of multi-intersection speed guidance in this embodiment;

[0099] Figure 9 This is an example diagram of breadth-first search in this embodiment;

[0100] Figure 10 This is the collaborative control flow chart of this embodiment;

[0101] Figure 11 This is a relationship diagram between the importance of test cases and the number of traffic conflicts in this embodiment;

[0102] Figure 12 Graphs showing average delay and average waiting time for the four control methods of this embodiment at different penetration rates. DETAILED DESCRIPTION

[0103] To make the technical solutions, advantages, and purposes of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0104] like Figure 1 As shown, the present invention provides a generalized test case generation method for multi-intersection IVCPS collaborative control, comprising the following steps:

[0105] S1. Construct a basic graph model of mixed traffic flow, combine the IDM and CACC models, and derive the relationship between traffic flow density and flow under different penetration rates;

[0106] S1.1 Construct a basic graph model of mixed traffic flow and calculate the average headway and flow-density relationship of mixed traffic flow;

[0107] The three traffic quantities in the traditional traffic flow basic graph model are traffic flow (q, unit: vehicles / hour), traffic density (k, unit: vehicles / km) and average speed (v, unit: km / h). The relationship between them is:

[0108] q=k·v

[0109] According to traffic flow theory, traffic flow density and headway distance at traffic flow equilibrium are reciprocals of each other. The traffic flow density and flow rate of HVs and CAVs can be obtained as follows:

[0110]

[0111] Where v represents the average speed of HV and CAV; k H represents the traffic density of HV; q H represents the flow rate of HV; v0 represents the expected speed of HV and CAV vehicles; s0 is the minimum safe distance between HV and CAV when they are stationary; T H represents the safe headway of HV; L represents the vehicle length; k C represents the traffic flow density of CAV; q C represents the flow rate of CAV; T C It represents the safe headway of CAV;

[0112] Assuming that the mixed traffic flow consists only of human-driven vehicles (HV) and connected and automated vehicles (CAV) with different levels of automation, the average headway h in the equilibrium state of the mixed traffic flow is derived. M The table is:

[0113] h M =p·h C +(1-p)·h H

[0114] Where p represents the penetration rate of intelligent connected vehicles in the total vehicle population; 1-p represents the penetration rate of human-driven vehicles in the total vehicle population; h H Indicates the distance between HV vehicles; h C Indicates the distance between CAV heads;

[0115] Combining the two equations, the basic graph model of mixed traffic flow is obtained as follows:

[0116]

[0117] Where k M represents the traffic density of mixed traffic flow; q M represents the flow rate of mixed traffic flow;

[0118] S1.2 Dynamically adjust the headway ratio between CAV and HV based on the penetration rate p;

[0119] In mixed traffic flows, one type is human-driven vehicles, whose following behavior is described by the IDM model; the other type is intelligent connected vehicles, whose following behavior is described by the ACC or CACC model. Assume that the penetration rate of intelligent connected vehicles in the total vehicle population is p (i.e., p vehicles use ACC / CACC, and the remaining 1-p vehicles use IDM). A basic graph model for the flow density of mixed traffic flows is constructed using the homogeneous traffic flow basic graph model. When the traffic flow is balanced, different types of vehicles maintain the same speed v, but the following distances they maintain vary. The equilibrium speed satisfies:

[0120]

[0121] Substituting the above formula into the core formula of IDM and CACC, we can get the headway h of the homogeneous traffic flow between human-driven vehicles and intelligent connected vehicles. H for and h C The table is as follows:

[0122]

[0123] h C =L+s0+T C v

[0124] S2. Based on the basic traffic flow graph model, expand the IVCPS scenario ontology and design multi-intersection road network scenarios, such as Figure 2-Figure 5 As shown;

[0125] S2.1 Based on the basic traffic flow graph model, the specific steps for analyzing the traffic flow model include:

[0126] For the human driving model, the Intelligent Driver Model (IDM) is selected to simulate the driver's acceleration and deceleration behavior during the car-following process to reproduce various traffic flow phenomena, such as free flow, synchronized flow, and congested flow. The core formula of the IDM is as follows:

[0127]

[0128] Among them, a n is the acceleration of vehicle n; a max is the maximum acceleration of the vehicle; v n is the current speed of the vehicle; v0 is the desired speed of the vehicle; s n is the actual distance between the vehicle and the preceding vehicle; s0 is the minimum safe distance when stationary; T H is the reaction time (safe headway); b is the comfortable deceleration; Δv n is the speed difference between the vehicle and the preceding vehicle; L is the length of the vehicle.

[0129] For smart cars, the Cooperative Adaptive Cruise Control (CACC) model is selected. The core formula of CACC is as follows:

[0130]

[0131] Among them, a i (t) is the acceleration of the vehicle at time t; a i-1 (t) is the acceleration of the preceding vehicle; e i (t) is the following error; k is the rate of change of the following error; p for and k d is the proportional-derivative control gain; p i (t) is the position of the vehicle at time t; p i-1 (t) is the position of the preceding vehicle at time t; L is the length of the vehicle; s0 is the minimum safe distance maintained when the vehicle is stationary; T C is the preset headway; v i (t) is the speed of the vehicle at time t.

[0132] Assuming that the mixed traffic flow consists of only HVs and CAVs, when the traffic flow is balanced, different types of vehicles maintain the same speed v, but the following distances they maintain are different. The equilibrium speed satisfies: all vehicles have the same speed and zero acceleration. At this time, the HV headway h H and CAV headway h C The calculation formulas are:

[0133]

[0134] h C =L+s0+T C v

[0135] The average headway distance of the mixed flow is:

[0136] h M =p·h C +(1-p)·h H

[0137] The basic graph model of mixed traffic flow is:

[0138]

[0139] S2.2 The specific steps for designing a multi-intersection road network scenario are as follows:

[0140] (1) Clarify design goals and scale conversion

[0141] First, the goal of scaling from a single intersection to a macroscopic road network is to clarify. Because microscopic parameters (such as vehicle trajectories) significantly increase complexity in large-scale scenarios, it is necessary to couple microscopic behaviors to macroscopic characteristics through traffic flow models (such as IDM / CACC). This requires focusing on overall metrics such as flow rate and density, while retaining key parameters such as permeability to quantify the mixed impact of CAVs and HVs.

[0142] (2) Expanding the IVCPS Scenario Ontology

[0143] Based on the IVCPS structured model, a new dynamic object layer (traffic flow entities and their attributes, such as volume and vehicle type ratio), a static road layer (road network topology and intersection connectivity), and a coordinated control layer (signal coordination parameters, such as phase difference and sub-area division) have been added. Through layered expansion, unified modeling of microscopic details and macroscopic characteristics has been achieved, providing a data foundation for subsequent control optimization.

[0144] (3) Constructing a road network model

[0145] The road network structure is represented by a directed graph G = (J, L) G = (J, L), where the intersection set JJ and the road set LL (including entry and exit roads) form the nodes and edges. Vehicle routing is based on the set of feasible paths Rj and relies on a two-tier architecture consisting of an edge cloud (real-time CAV control) and a regional cloud (global path optimization) to achieve dynamic resource allocation and collaborative management.

[0146] (4) Design collaborative control mechanism

[0147] Signal coordination is achieved through sub-area division and phase difference optimization, dynamically adjusting green light duration. CAVs utilize V2X communication for route and speed guidance, while HVs rely on traditional sensors. Cloud-edge collaboration (real-time response from the edge cloud and global scheduling from the regional cloud) ensures improved efficiency at the road network level.

[0148] (5) Verification and testing

[0149] System performance is verified through a combination of test cases (covering parameters such as penetration rate and vehicle model ratio), using both macro-level (delay, throughput) and micro-level (ETTC / PET conflict statistics). Test results are fed back into model optimization, forming a closed-loop design process.

[0150] S3. A two-layer distributed collaborative control architecture is used: the upper layer dynamically adjusts signal timing based on traffic pressure information, and the lower layer provides optimal path and driving speed recommendations for CAVs. The collaborative control process is as follows: Figure 10 shown.

[0151] In the upper-level signal control optimization of S3.1, estimate the intersection queue length;

[0152] The estimated queue length at a signalized intersection is defined as follows:

[0153] q i,o (t+1)=[q i,o (t)-α i,o (t)·(s i,o (t)-l i,o (t))+d i,o (t)+a i,o (t)] +

[0154] Among them, q i,o (t+1) is the road L in a certain direction of the intersection i , and L o →L i is the estimated value of the queue length at the next moment in the driving direction; q i,o (t) is the current queue length; α i,o (t) is the release coefficient at time t in that direction; s i,o (t) is the saturated flow rate of the road, which is determined by factors such as road width, number of lanes, saturated flow rate, signal timing, etc.; i,o (t) is the flow loss; d i,o (t) planning the vehicle flow that will pass through this road section and arrive soon for other routes in the road network; a i,o (t) is the vehicle flow directly arriving at the road from outside the road network, and is set to 0 when external traffic is not considered; [·] + For positive operations, when the result is negative, it is recorded as 0 because the queue length cannot be negative.

[0155] S3.2 dynamically adjusts the phase green light duration;

[0156] Taking queue length as the metric in the traffic pressure formula, we get the core formula of traffic pressure:

[0157]

[0158] Among them, q j,i For intersection L i →L j Direction Road L j The queue length of road L j For road L i Downstream of; in phase I road L i There are multiple downstream roads, phaseI is the set of these downstream roads; i,o For L i →L j steering ratio.

[0159] The corresponding traffic pressure can be obtained for all phases of the intersection, and the traffic lights are controlled using a cyclic control method based on the traffic pressure. Based on a fixed control cycle and phase sequence, the green light time is allocated according to the proportion of traffic pressure in all phases. The allocation formula is as follows:

[0160]

[0161] Among them, g i is the green light time of phase i; C is the signal period; p N is the number of signal phases; g min The minimum green light time.

[0162] In the lower-level control of S3.3, a breadth-first search algorithm is used to optimize path selection;

[0163] The directed graph of the urban area road network is G = (J, L), where J is the set of intersection nodes, J0 is the root node, and J d is the destination node. L is the road set, each road L i (Direction is determined) with attribute road length l(L). Calculate the expected queue length in the search direction, t e is the expected arrival time estimated by the sum of the current speed and the road length on the current backtracking path, t is the current moment, max(q L (t),Q m ) is the queue length q at the current moment L (t) The maximum length Q of traffic that can pass at the same time as the green light in that direction m The larger value of d L (t e ) is the length of the traffic flow that will arrive at the intersection near the expected time in the path planning.

[0164] The cost increment for each downward search step of the path is defined as follows:

[0165]

[0166] Where γ is the relative importance weight, which is used to balance the estimated queue length and road length.

[0167] For a path R (consisting of a series of continuous roads and intersections), the total cost function is defined as follows:

[0168]

[0169] The goal of optimizing the search is as follows:

[0170]

[0171] Indicates the path from the root node J0 to the destination node J dAmong the feasible paths, select the total cost The minimum path. The breadth-first search algorithm starts from the root node. Each time the node expands downward, the cost increment of the edge is calculated and the cumulative cost from the root node to the current node is updated. When the search reaches the maximum depth condition, the search ends. The candidate path search example is as follows Figure 9 .

[0172] S3.4 Calculation of coordinated speed guidance at multiple intersections:

[0173] According to the distance d from the i-th intersection ahead i , the current signal phase and remaining time, vehicle position, speed, acceleration, driving history data and other factors to dynamically adjust the vehicle speed. When the vehicle enters the edge cloud speed guidance range, the time when the vehicle travels at a constant speed v0 at time t0 to the stop line of the i-th intersection is predicted as follows:

[0174]

[0175] The green light period of each intersection can be expressed as a time interval [g start,i ,g end,i ], the vehicle should ensure that its arrival time falls within this interval, requiring:

[0176] g start,i ≤(t i -φ i )modC i ≤g end,i

[0177] Among them, C i is the signal period of intersection i, φ i is the phase offset of the intersection (i.e., the absolute or relative reference time of the green light start time), and the signal periodicity is considered using modular operation.

[0178] In order to take into account multiple intersections, a constraint set is constructed:

[0179]

[0180] Solving this set of inequalities yields a speed range that satisfies all green light conditions. If multiple feasible speeds exist, the system typically selects the one that optimizes driving comfort and traffic flow stability, taking into account safety margins (e.g., compensating for reaction time Δt). The recommended speed for each intersection is:

[0181]

[0182] Where, t g,iIndicates the remaining green light time of the current phase of the intersection when the vehicle arrives at the i-th intersection (unit: seconds);

[0183] Multi-intersection speed guidance Figure 8 shown.

[0184] S4. Build a scenario parameter extraction module to extract control algorithms, vehicle type ratios, traffic flow, and CAV penetration rates:

[0185] S4.1 Extraction control algorithm;

[0186] (1) Fixed timing

[0187] The core parameters of fixed timing include cycle length, phase allocation, and green-to-signal ratio. The optimal cycle length can be calculated using the Webster formula as follows:

[0188]

[0189] The green light duration of each phase is calculated as follows:

[0190]

[0191] Where L is the total loss time in each cycle (including the loss time of starting, reaction, yellow flash time, etc.), Y is the sum of the ratio of each phase flow rate to the saturation flow rate, v i and s i are the flow rate and saturation flow rate of the i-th phase respectively.

[0192] (2) Induction timing control

[0193] The basic principle is to first operate at a fixed minimum timing. When a vehicle comes from a certain direction, the green light time is appropriately extended. Finally, the phase transition is determined by judging whether the maximum green light time has been reached and whether a vehicle has been detected. The flow chart is as follows: Figure 6 .

[0194] The important parameters of the system include the preset minimum green light duration T for each phase min ; After a vehicle is detected during the extended detection period, the fixed time T added to each extension of the green light ext ; Determine whether the time interval T of vehicles arriving continuously gap When no vehicle is detected within this period, the current green light ends; to prevent the green light time in one direction from being too long and affecting the passage of other roads, the upper limit of the green light time T is set. maxTherefore, the effective green light duration of the full-sensor timing control is the minimum green light duration plus the extended duration. The maximum green light time and the minimum green light time are generally extended and shortened by 20% to 50% of the fixed green light time. The extension increment depends on the response of the detection equipment and the vehicle interval time. It is usually set to 3 to 5 seconds. It can respond to the arrival of subsequent vehicles in a timely manner and avoid frequent switching due to too short a time. The schematic diagram of important parameters is shown as follows Figure 7 .

[0195] S4.2 Extract vehicle model ratio;

[0196] The extraction of vehicle type ratios aims to quantify the impact of different vehicle types (such as small cars and large cars) on traffic flow characteristics. First, structured attribute data is obtained from the traffic flow entities in the dynamic object layer, including the proportion of small cars and the proportion of large vehicles (buses, heavy vehicles, etc.). These data are collected in real time by roadside sensing equipment (such as cameras and lidar) and calibrated with vehicle registration information. Secondly, vehicle dynamic parameters need to be associated: large vehicles have longer body lengths and greater deceleration, so their following distance is significantly larger than that of small vehicles, which directly affects the calculation of traffic flow density. In the experimental design, typical ratios are set to 0:1 (pure small cars), 0.1:0.9 and 0.2:0.8 to cover mixed traffic scenarios from light to heavy.

[0197] S4.3 Extract traffic flow;

[0198] Traffic flow extraction is used to evaluate the load level of the road network, covering the entire scenario from free flow to congestion. The core data comes from the flow q (vehicles / hour) and density k (vehicles / km) collected in real time by road detectors (such as induction coils and radars), and the data consistency is verified by the basic traffic flow graph model. The saturated flow is further combined with static parameters such as the number of lanes and signal timing to dynamically correct the road capacity. During the experimental verification phase, the flow rate is divided into four levels: 2000 veh / h (free flow), 4000 veh / h (stable flow), 6000 veh / h (peak congestion), and 8000 veh / h (severe congestion), corresponding to the test requirements of different control strategies.

[0199] When extracting traffic flow, the contributions of CAVs and HVs are distinguished. CAVs provide precise speed and position data due to their communication capabilities, while HV traffic flow must be estimated indirectly through fixed sensors. In the mixed flow model, traffic flow data is linked to the penetration rate p. For example, at high penetration rates, the small headway between CAVs can increase the maximum value.

[0200] S4.4 Extract CAV permeability;

[0201] The penetration rate is defined as the ratio of CAVs to the total number of vehicles. The data sources include direct statistics of V2X communication and edge cloud fusion perception results (such as camera + DSRC). In the hybrid flow model, p dynamically adjusts the headway weight: CAV adopts the CACC model, and HV adopts the IDM model, resulting in h C <h H The experimental setting covers five penetration rates: 0% (pure HV), 25%, 50%, 75%, and 100% (pure CAV) to analyze the adaptability of the control method.

[0202] The vehicle behavior is updated and correlated in real time when the penetration rate is extracted. For example, when p>50%, cooperative control significantly reduces delays through global path optimization of CAVs, such as Figure 12 At the same time, changes in penetration affect traffic conflict indicators (ETTC / PET). For example, CAV coordinated lane changing can reduce the number of conflicts in high-p scenarios. The module dynamically calibrates the p value to ensure that test cases cover all stages of technology evolution.

[0203] S5. Build a key calculation module to calculate the key indicators of parameters;

[0204] ETTC and PET are used to count traffic conflicts. Enhanced Time-to-Collision (ETTC) and Post-Encroachment Time (PET) are used as indicators to count the number of traffic conflicts within the road network. ETTC, an improvement on traditional TTC, considers relative acceleration when calculating collision time, which is more advantageous than TTC and can better assess the risk level of the scene. Its calculation formula is as follows:

[0205]

[0206] Among them, Ra is the relative distance between the two vehicles; Δv is the relative speed; Δa is the relative acceleration.

[0207] PET is defined as the time interval between the first vehicle leaving the conflict area and the next vehicle entering the conflict area. The calculation formula is as follows:

[0208] PET=t f -t b

[0209] Where, t f The time when the vehicle that passes first leaves the conflict area; t b The time when the last vehicle enters the conflict area;

[0210] The relationship between the importance of test cases and the number of traffic conflicts is as follows: Figure 11 shown.

[0211] S6. Build a test case generation module and use the AETG-S algorithm to generate a set of combined test cases;

[0212] The control algorithm, vehicle type ratio (small cars, large cars), traffic flow, and CAV penetration rate are selected as scenario parameters, and the relative criticality between them is calculated to represent the relative criticality between the generated test cases. The relative criticality index between the values ​​of the four traffic flow elements is calculated as follows:

[0213]

[0214] in, is the scene parameter; Importance is a subjective score given by experts based on their experience, expertise, and understanding of scenario concepts; Exposure frequency reflects the probability of scene concept attributes appearing in actual road scenes, which can be calculated through naturalistic driving data (NDD) analysis; is the scene parameter A specific value of Ω T is the set of coordinates of all concept nodes on the backtracking path T; w ij is the relative importance of the corresponding coordinate concept node; Indicates that within the observation period (or space), the scene element value is The total time (or frequency) of the table; f total Represents the total time (or frequency) of the observation period (or space).

[0215] With coverage strength t=2, 33 test cases were generated using the AETG-S algorithm. These test cases cover all 2-way combinations of the four traffic flow attributes, and the number of test cases is greatly reduced compared to the 240 test cases for the full combination.

[0216] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of protection of the present invention.

Claims

1. A generalized test case generation method for multi-intersection IVCPS collaborative control, characterized by: The following steps are involved: S1. Construct a basic graph model of mixed traffic flow, combine the IDM and CACC models, and derive the relationship between traffic flow density and flow under different penetration rates; S2. Based on the basic graph model of mixed traffic flow, expand the IVCPS scenario ontology and design a multi-intersection road network scenario; S3. A two-layer distributed collaborative control architecture is used: the upper layer dynamically adjusts signal timing based on traffic pressure information, while the lower layer provides optimal route and speed recommendations for CAVs. S4. Build a scenario parameter extraction module to extract control algorithms, vehicle type ratios, traffic flow, and CAV penetration rates: S5. Build a key calculation module to calculate the key indicators of parameters; S6. Build a test case generation module and use the AETG-S algorithm to generate a combination test case set.

2. A method for generalizing test cases for multi-intersection IVCPS collaborative control according to claim 1, characterized in that: The step S1 includes the following sub-steps: S1.1 Construct a basic graph model of mixed traffic flow and calculate the average headway and flow-density relationship of mixed traffic flow; The traffic flow density and flow rate of HV and CAV are: Where v represents the average speed of HV and CAV; k H represents the traffic density of HV; q H represents the flow rate of HV; v0 represents the expected speed of HV and CAV vehicles; s0 is the minimum safe distance between HV and CAV when they are stationary; T H represents the safe headway of HV; L represents the vehicle length; k C represents the traffic flow density of CAV; q C represents the flow rate of CAV; T C It represents the safe headway of CAV; Assuming that the mixed traffic flow consists only of human-driven vehicles HV and intelligent connected vehicles CAV, the average headway h in the equilibrium state of the mixed traffic flow is derived: M , h M The calculation expression is: h M =p·h C +(1-p)·h H Where p represents the penetration rate of intelligent connected vehicles in the total vehicle population; 1-p represents the penetration rate of human-driven vehicles in the total vehicle population; h H Indicates the distance between HV vehicles; h C Indicates the distance between CAV heads; The hybrid traffic flow basic graph model is constructed by the isomorphic traffic flow basic graph model. The hybrid traffic flow basic graph model is: Where k M represents the traffic density of mixed traffic flow; q M represents the flow rate of mixed traffic flow; S1.2 Dynamically adjust the headway ratio between CAV and HV based on the penetration rate p; The IDM model is used to describe the following behavior of human-driven vehicles, and the ACC or CACC model is used to describe the following behavior of intelligent connected vehicles. When the traffic flow is balanced, different types of vehicles maintain the same driving speed v, but the following distances maintained by human-driven vehicles and intelligent connected vehicles are different. The equilibrium speed satisfies: Substituting the above formula into the core formula of IDM and CACC, we can get the HV headway h H Distance from CAV front h C The calculation expression is: h C =L+s0+T C v。 3. A method for generalizing test cases for multi-intersection IVCPS collaborative control according to claim 2, characterized in that: In step S2, the design of the multi-intersection road network scenario includes the following sub-steps: I. Clarify design goals and scale conversion; Clarify the expansion goal from a single intersection to a macro road network, and couple microscopic behaviors to macroscopic characteristics through the traffic flow model IDM / CACC; II. Expanding IVCPS scenario ontology; Based on the IVCPS structured model, a new dynamic object layer, static road layer, and coordination control layer are added. Through layered expansion, unified modeling of microscopic details and macroscopic characteristics is achieved; III. Construct road network model; The road network structure is represented by a directed graph G = (J, L), where the intersection set JJ and the road set LL containing entrances and exits constitute nodes and edges. Vehicle path planning is based on the feasible path set Rj and relies on a two-layer architecture of edge cloud and regional cloud to achieve dynamic resource allocation and collaborative management. IV. Design collaborative control mechanisms; Signal coordination is achieved through sub-area division and phase difference optimization, dynamically adjusting green light duration. CAVs utilize V2X communication for route and speed guidance, while HVs rely on traditional sensors. V. Verification and testing; System performance is verified through combined test cases, with evaluation indicators including macro and micro. Test results are fed back to model optimization to form a closed-loop design process.

4. A method for generalizing test cases for multi-intersection IVCPS collaborative control according to claim 3, characterized in that: In step S2, for human driving, an intelligent driver model (IDM) is selected to simulate the driver's acceleration and deceleration behaviors during the car-following process to reproduce various phenomena in traffic flow. The core formula of the IDM is: Where a n is the acceleration of vehicle n; a max is the maximum acceleration of the vehicle; v n is the current speed of the vehicle; v0 is the desired speed of the vehicle; s n is the actual distance between the vehicle and the preceding vehicle, s0 is the minimum safe distance when stationary, T H is the safe headway; b is the comfortable deceleration; Δv n is the speed difference between the vehicle and the preceding vehicle; L is the vehicle body length.

5. The method for generalizing test cases for multi-intersection IVCPS collaborative control according to claim 4 is characterized in that: In step S2, a cooperative adaptive cruise control (CACC) model is selected for the smart car. The core formula of CACC is: Where a i (t) is the acceleration of the vehicle at time t; a i-1 (t) is the acceleration of the preceding vehicle; e i (t) is the following error; k is the rate of change of the following error; p and k d is the proportional-derivative control gain; p i (t) is the position of the vehicle at time t; p i-1 (t) is the position of the preceding vehicle at time t; L is the vehicle length; s0 is the minimum safe distance when the vehicle is stationary; T C is the preset headway; v i (t) is the speed of the vehicle at time t.

6. A method for generalizing test cases for multi-intersection IVCPS collaborative control according to claim 5, characterized in that: The step S3 includes the following sub-steps: S3.1 Estimate the length of the queue at the intersection; The estimated queue length at a signalized intersection is defined as follows: q i,o (t+1)=[q i,o (t)-α i,o (t)·(s i,o (t)-l i,o (t))+d i,o (t)+a i,o (t)] + Where q i,o (t+1) is the road L in a certain direction of the intersection i , and L o →L i is the estimated value of the queue length at the next moment in the driving direction; q i,o (t) is the current queue length; α i,o (t) is L o →L i Release coefficient at time t in the direction; s i,o (t) is the road L i Saturation flow; l i,o (t) is the flow loss; d i,o (t) planning the vehicle flow that will pass through this road section and arrive soon for other routes in the road network; a i,o (t) is the road L that can be directly reached from outside the road network i Vehicle flow, set to 0 when external flow is not considered; [·] + For positive operations, when the result is negative, the result is recorded as 0, because the queue length cannot be negative; S3.2 dynamically adjusts the phase green light duration; Taking queue length as the metric in the traffic pressure formula, we get the core formula of traffic pressure: Where q j,i For intersection L i →L j Direction Road L j The queue length of road L j For road L i Downstream of; in phase I road L i There are multiple downstream roads, phaseI is the set of downstream roads; λ i,o For L i →L j Steering ratio; Traffic lights are controlled using a cyclic control method based on traffic pressure. Based on a fixed control cycle and phase sequence, green light time is allocated according to the proportion of traffic pressure in all phases. The allocation formula is as follows: Where g i is the green light time of phase i, C is the signal period, p N is the number of signal phases, g min is the minimum green light time; S3.3 uses a breadth-first search algorithm to optimize path selection; The cost increment for each downward search step of the path is defined as follows: Where γ is the relative importance weight, which is used to balance the estimated queue length and road length; q L is the queue length at the current moment; l(L) is the road length; For a path R, the total cost function is defined as follows: The goal of optimizing the search is as follows: Indicates the path from the root node J0 to the destination node J d Among the feasible paths, select the total cost The smallest path; S3.4 Recommendations for coordinated speed guidance at multiple intersections; Usually, a speed that is optimal in terms of driving comfort and traffic flow stability is selected, and safety margins are considered. The recommended speeds for each intersection are as follows: Where, d i is the distance to the i-th intersection ahead; t g,i It represents the remaining green light time of the current phase of the i-th intersection when the vehicle arrives at the i-th intersection; Δt represents the compensation reaction time.

7. The method for generalizing test cases for multi-intersection IVCPS collaborative control according to claim 6 is characterized by: The key indicators of the calculation parameters in step S5 include enhanced time to collision (ETTC) and post-entry time (PET); The calculation expression of ETTC is: Where Ra is the relative distance between the two vehicles; Δv is the relative velocity; Δa is the relative acceleration; The calculation expression of PET is: PET=t f -t b Where, t f The time it takes for the vehicle that passes first to leave the conflict area; t b The time when the last vehicle enters the conflict area.

8. The method for generalizing test cases for multi-intersection IVCPS collaborative control according to claim 7 is characterized in that: The specific content of step S6 is: The control algorithm, vehicle type ratio, traffic flow, and CAV penetration rate are selected as scenario parameters. The relative criticality between the parameters is calculated to represent the relative criticality between the generated test cases. The relative criticality index between the values ​​of the four traffic flow elements is obtained by calculation, and the formula is as follows: Where, is the scene parameter; For importance; The exposure frequency reflects the probability of the scene concept attribute appearing in the actual road scene; is the scene parameter Is a specific value; Ω T is the set of coordinates of all concept nodes on the backtracking path T; w ij is the relative importance of the corresponding coordinate concept node; Indicates that the scene element value during the observation period is The total time of the scene element in the observation space is The total frequency of f total Indicates the total time of the observation period or the total frequency of the observation space; With coverage strength t=2, the AETG-S algorithm is used to generate a set of combined test cases.

Images