Long queue hierarchical control model method under multi-virtual head vehicle structure

Abstract

The application discloses a long queue hierarchical control model method under a multi-virtual-head vehicle structure, and comprises the following steps: collecting long queue vehicle driving data by a roadside control end; loading the driving data into an optimal virtual-head vehicle sequence calculation model; determining whether the model output result meets a genetic algorithm termination condition; if yes, outputting the optimal virtual-head vehicle sequence; if not, continuing to perform selection operation, crossover operation and mutation operation in sequence; and taking the optimal virtual-head vehicle sequence as a basis, introducing a speed difference between adjacent sub-queue head vehicles of the long queue and a speed difference between a sub-queue following vehicle and a sub-queue head vehicle into control layers respectively to obtain a sub-queue head vehicle control layer model and a sub-queue following vehicle control layer model. The built model has high recognition degree for head vehicle speed change, can effectively reduce the speed disturbance influence caused by speed change, and ensures the high efficiency of long queue driving of the automatic driving vehicle.

Description

Technical Field

[0001] This invention relates to the field of intelligent vehicle application technology, specifically to a method for a hierarchical control model of a long queue under a multi-virtual lead vehicle structure. Background Technology

[0002] The development of intelligent connected vehicles has been strengthened, and autonomous driving technology based on vehicle-road cooperation has received widespread attention. Leveraging emerging C-V2X communication technology, platooning on highways is gradually moving towards practical application, with control methods including Adaptive Cruise Control (ACC) and Cooperative Adaptive Cruise Control (CACC). Based on the cooperation of sensors and controllers, vehicles within the platoon share information in real time through communication, ensuring the stability of platooning. Autonomous vehicle platooning on highways can effectively improve road traffic efficiency and reduce vehicle fuel consumption. Autonomous vehicle platooning mainly involves merging and controlling vehicles traveling in the same direction, ensuring that vehicles in the same group maintain a consistent speed and that adjacent vehicles within the platoon maintain a desired distance. Effective control of platoon spacing improves road capacity, reduces air resistance for following vehicles, and thus reduces energy consumption. Existing research suggests that platooned vehicles can reduce fuel consumption by nearly 5% compared to individual vehicles; therefore, platooning control strategies for autonomous vehicles are more economically efficient than single-vehicle control strategies. With the development of C-V2X communication technology, long convoy driving on highways has become a new development direction. However, when long convoys of autonomous vehicles travel on the main road of a highway, a speed change in the lead vehicle can rapidly affect all following vehicles, causing them to experience speed fluctuations. The convoy can only return to a stable, uniform speed state once the speed disturbance completely dissipates. During the period of speed disturbance, the speeds of the following vehicles in the convoy are unstable, and due to these repeated speed fluctuations, the vehicles consume more energy.

[0003] Existing queue control methods are all constrained by communication range, often ensuring queue reliability by dividing long queues into multiple sub-queues, with each sub-queue controlled and guided by a virtual lead vehicle. Furthermore, current research mainly focuses on improving vehicle communication capabilities within the communication network built on the virtual lead vehicle, with relatively little research addressing the optimal distribution of virtual lead vehicles within long queues for autonomous vehicles. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a hierarchical control model method for long queues under a multi-virtual lead car structure.

[0005] The present invention adopts the following technical solution:

[0006] A hierarchical control model method for long queues under a multi-virtual lead car structure includes the following steps:

[0007] Step 1: The roadside control unit collects driving data of long queues of vehicles;

[0008] Step 2: Load the driving data described in Step 1 into the optimal virtual lead vehicle sequence calculation model;

[0009] Step 3: Check if the output of the model in Step 2 meets the termination condition of the genetic algorithm. If it does, output the optimal virtual lead car sequence; if not, continue to perform selection, crossover, and mutation operations in sequence.

[0010] Step 4: The vehicle platoon control model is based on the optimal virtual lead vehicle sequence in Step 3. The speed difference between the lead vehicles of adjacent sub-platoons and the speed difference between the following vehicles and the lead vehicles of the sub-platoons are introduced into the control layer to obtain the sub-platoon lead vehicle control layer model and the sub-platoon follow vehicle control layer model. By synchronously controlling the sub-platoon lead vehicles and the following vehicles, the operating status of all autonomous vehicles in the long platoon is updated and used for the calculation of the optimal virtual lead vehicle sequence in the next time step.

[0011] Furthermore, the optimal virtual lead vehicle sequence calculation model is as follows:

[0012]

[0013]

[0014]

[0015]

[0016] In the formula C i efficiency Let be the efficiency cost of the i-th following vehicle, Ci string be the stability cost of the i-th following vehicle, v0 t be the speed of the lead vehicle in the long queue at time t, a0 t be the acceleration of the lead vehicle in the long queue at time t, vit be the speed of the i-th following vehicle in the long queue at time t, and a t i Let γ be the acceleration of the i-th following vehicle in the long queue at time t, and T be the safe headway of the queue, which takes a value of 6 time steps, i.e., 0.6s. i For the attribute state value of the lead car in the subqueue, when γ i When γ equals 1, the i-th following car in the queue has the attribute of being the head car of the sub-queue. iThe i-th following vehicle in the queue with a duration not equal to 1 does not have the attribute of the lead vehicle in the subqueue. N is the total number of autonomous vehicles, and n is the total number of virtual lead vehicles.

[0017] Furthermore, the specific steps of the genetic algorithm in step 3 are as follows:

[0018] Step 3.1: Initialization, using the attributes of the first car in the sub-queue within the long queue as the solution variable, constructing individuals using binary encoding, generating an initial population that meets the requirements, and determining the algorithm parameters;

[0019] Step 3.2: Greedy repair. Encode and repair individuals that do not meet the model constraints. If the sub-queue head car attribute is assigned to multiple consecutive cars in the result, the sub-queue vehicle quantity constraint is not valid. Compare the cost values ​​of each car that does not meet the constraint, retain the car with the best cost value among the adjacent cars as the sub-queue head car, and cancel the sub-queue car attribute of the adjacent cars so that the solution result meets the model constraints.

[0020] Step 3.3: Individual assessment: Calculate the fitness of each individual based on the information of all individuals in the population;

[0021] Step 3.4: Determine the generation number. If the generation number has not reached the termination generation number, perform the genetic operations from Step 3.4 to Step 3.6 on the population. If the termination generation number is reached, select the individual with the best fitness performance in the current population as the optimal solution and end the solution for the optimal virtual lead car sequence at this time step.

[0022] Step 3.5: Selection operation. Based on the fitness shown by all individuals in the population, superior individuals are selected through roulette wheel selection and retained for the next generation.

[0023] Step 3.6: Crossover operation. In the current population, chromosomes of two individuals are exchanged in pairs at random crossover locations using a single-point crossover method, resulting in new individuals entering the population.

[0024] Step 3.7: Mutation operation, which alters some genes in all individuals of the population with a mutation probability to produce entirely new individuals. After the genetic operation, the next generation of the population re-enters the evaluation stage of Step 3.2.

[0025] Furthermore, the model of the sub-queue lead vehicle control layer is as follows:

[0026]

[0027] In the formula, j is the sequential position of the virtual lead car in the long queue, ei t is the deviation between the actual distance and the expected distance between the lead car i and the preceding car i-1 in the sub-queue, ai t+1 is the acceleration of the lead car i in the sub-queue at time t+1, vi t is the velocity of the lead car i in the sub-queue at time t, vi-1t is the velocity of the preceding car i-1 in the sub-queue at time t, vj t is the velocity of the adjacent lead car j in the sub-queue at time t, k0 is the acceleration weight of the preceding car in the sub-queue lead car control model, k1 is the distance error weight in the sub-queue lead car control model, k2 is the speed difference weight of the preceding car in the sub-queue lead car control model, k3 is the speed difference weight between the adjacent lead car in the sub-queue and the long queue lead car in the sub-queue lead car control model, and k4 is the speed difference weight of the adjacent lead car in the sub-queue that does not include the long queue lead car.

[0028] Furthermore, the model of the sub-queue following vehicle control layer is as follows:

[0029]

[0030] In the formula, ei t is the deviation between the actual distance and the expected distance between the following vehicle i and the preceding vehicle i-1 in the sub-queue, ait+1 is the acceleration of the following vehicle i in the sub-queue at time t+1, vi t is the velocity of the following vehicle i in the sub-queue at time t, vi-1t is the velocity of the preceding vehicle i-1 in the sub-queue at time t, vj t is the velocity of the leading vehicle j in the sub-queue to which the following vehicle belongs at time t, K0 is the acceleration weight of the preceding vehicle in the sub-queue following vehicle control model, K1 is the distance error weight in the sub-queue following vehicle control model, K2 is the speed difference weight of the preceding vehicle in the sub-queue following vehicle control model, and K3 is the speed difference weight of the leading vehicle in the sub-queue to which the following vehicle belongs in the sub-queue following vehicle control model.

[0031] The beneficial effects of this invention are:

[0032] (1) The long queue control model for autonomous vehicles, which introduces the speed difference parameters between adjacent lead vehicles in the sub-queue and the speed difference parameters between following vehicles and lead vehicles in the sub-queue, effectively reduces the impact of speed disturbances caused by changes in the lead vehicle's speed, making the speed distribution of vehicles in the long queue more convergent and the acceleration process of the queue vehicles smoother. The established model can reasonably respond to the speed fluctuations of the lead vehicle in various acceleration and deceleration scenarios. All vehicles in the long queue can adapt to the speed disturbances caused by changes in the lead vehicle's speed within 20 seconds and achieve re-stabilization of the operating state within 360 seconds, effectively improving the adverse effects caused by speed fluctuations in the control of long queue vehicles.

[0033] (2) The vehicles in the sub-queues divided by the virtual lead car within the long queue did not exhibit an even distribution, indicating that the existing uniform distribution scheme for sub-queues did not consider the influence of the current driving state of the queue vehicles. Secondly, when the speed of the lead car changes, the established model obtains the optimal solution to adapt to the current speed change at the same time, and the established model has a strong ability to recognize the speed change of the lead car.

[0034] (3) Compared with the existing control model, the model established in this paper effectively controls the speed oscillation of the following vehicles, reduces the speed disturbance of the following vehicles in the long queue, and reduces the average energy consumption of the vehicles in the long queue. The average energy consumption of the queue vehicles is reduced by 17.00% within 60 seconds after the lead vehicle decelerates rapidly.

[0035] (4) When the total number of vehicles in a long queue increases, the initial speed oscillation of the following vehicles in the long queue increases by a smaller margin, and the average delay time for the vehicles in the long queue to complete acceleration is shorter. When the speed change rate of the lead vehicle in the long queue is greater, the speed disturbance effect is more severe. The speed disturbance has the greatest effect 10 seconds after the lead vehicle changes speed. The speed disturbance caused by the acceleration behavior of the lead vehicle has a stronger impact than that caused by the deceleration behavior, but the speed disturbance dissipates faster during the acceleration behavior. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments will be briefly described below. Obviously, the drawings described below only relate to some embodiments of the present invention and are not intended to limit the present invention.

[0037] Figure 1 This is a schematic diagram of the hierarchical control model framework for long queues in this invention;

[0038] Figure 2 a) The change in average speed of the queue; b) The change in the optimal virtual lead car sequence of the queue;

[0039] Figure 3 This is a comparison chart of the standard deviation of the speed of vehicles in a long convoy under the acceleration state of the lead vehicle in this invention;

[0040] Figure 4 This is a comparison chart of the standard deviation of the speed of vehicles in a long convoy under the deceleration state of the lead vehicle in this invention;

[0041] Figure 5 This is a comparison chart of the standard deviation of the speed of vehicles in a long convoy under the condition of the lead vehicle rapidly decelerating.

[0042] Figure 6 A comparison chart of the average speeds of vehicles in a long queue according to this invention;

[0043] Figure 7 A comparison chart of the standard deviations of long platoons of vehicles under acceleration conditions according to this invention;

[0044] Figure 8 Comparison chart of standard deviations of long platoons of vehicles under deceleration conditions according to this invention Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0046] Unless otherwise defined, the technical or scientific terms used in this disclosure shall have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "comprising" or "including," and similar words used in this disclosure, mean that an element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, but do not exclude other elements or objects. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0047] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0048] like Figure 1 As shown, a hierarchical control model method for long queues under a multi-virtual lead car structure includes the following steps:

[0049] Step 1: The roadside control terminal collects driving data of long queues of vehicles.

[0050] Step 2: Load the driving data described in Step 1 into the optimal virtual lead vehicle sequence calculation model.

[0051] The optimal virtual lead vehicle sequence calculation model, as the upper-level model, considers an objective function that encompasses both the driving efficiency and stability of the long platoon of autonomous vehicles. The driving efficiency is represented by the deviation of the virtual lead vehicle's speed from the lead vehicle's speed at a specified time step; the closer the virtual lead vehicle's speed is to the lead vehicle's speed, the higher the driving efficiency of the long platoon. The stability of the long platoon is represented by the cumulative acceleration damping ratio between the virtual lead vehicle and the lead vehicle within a specified time step. Based on the vehicle states during the long platoon's operation, two evaluation indicators are derived for the computational cost function of the optimal virtual lead vehicle sequence calculation model: the deviation between the virtual lead vehicle's speed and the lead vehicle's speed, and the long platoon stability index. The first indicator is calculated by the ratio of the squared speed difference between the virtual lead vehicle and the lead vehicle at that time step to the squared speed of the lead vehicle. The second indicator is calculated by the cumulative acceleration damping ratio between the virtual lead vehicle and the lead vehicle over a time period of length T before that time step. The calculation formula is as follows:

[0052]

[0053]

[0054] In the formula, C i efficiency Let C be the efficiency cost value of the i-th following vehicle. i string Let v be the stability cost value of the i-th following vehicle. 0 t Let a be the speed of the lead car in the long convoy at time t. 0 t Let v be the acceleration of the lead car in the long convoy at time t. i t Let a be the speed of the i-th following vehicle in the long queue at time t. i t Let t be the acceleration of the i-th following vehicle in the long queue at time t, and T be the safe headway of the queue, which is 6 time steps, i.e. 0.6s.

[0055] The cost function is determined by combining the two indicators, and its calculation formula is as follows:

[0056]

[0057] In the formula, Z virtual t γ represents the cost value of the optimal virtual lead car sequence obtained by the model at time t. i For the attribute state value of the lead car in the subqueue, when γ i When γ equals 1, the i-th following car in the queue has the attribute of being the head car of the sub-queue. iThe i-th following vehicle in the queue with a duration not equal to 1 does not have the attribute of the lead vehicle in the subqueue. N is the total number of autonomous vehicles, and n is the total number of virtual lead vehicles.

[0058] The core of a long queue of autonomous vehicles lies in the queue itself. A long queue structure requires at least 6 autonomous vehicles in its queue. Therefore, the total number N of vehicles in the queue should satisfy the following:

[0059] N≥6 (4)

[0060] There are N autonomous vehicles in a long queue, among which n are virtual lead vehicles with the attribute of being the lead vehicle in a sub-queue. This sub-queue lead vehicle attribute is defined using 0-1 variables; that is, when γi equals 1, the i-th following vehicle in the long queue has the sub-queue lead vehicle attribute. This allows us to obtain the state change information of the lead vehicle in the long queue, thus leading to the conclusion that:

[0061]

[0062] The lead car in a long queue, as the head car of the entire queue, also possesses the attributes of a sub-queue lead car. It can be used as the first virtual lead car, and its corresponding sub-queue lead car attribute γ0 is always equal to 1.

[0063] γ0=1 (6)

[0064] Combining equations (4), (5), and (6), we can obtain the following: The total number of virtual lead cars n should satisfy:

[0065] 1≤n<N (7)

[0066] Under the vehicle queuing structure constraint, the number of vehicles in each subqueue should be greater than or equal to 3. Therefore, the number of vehicles in each subqueue within a long queue of autonomous vehicles should be greater than or equal to 3.

[0067]

[0068] To ensure the constraint of equation (8), both the first and last sub-queues within the long convoy of autonomous vehicles must ensure that there are two or more following vehicles after the lead vehicle in the sub-queue. Therefore, these two sub-queues should satisfy:

[0069] γ i =0,1≤i<3 (9)

[0070] γ i =0, N-3<i≤N-1 (10)

[0071] Within a long convoy of autonomous vehicles, except for the first and last sub-convoys, all other sub-convoys must ensure that there are at least two following vehicles behind the lead vehicle in each sub-convoy; that is, the two virtual lead vehicles must be spaced at least two following vehicles apart.

[0072]

[0073] In summary, the optimal virtual lead vehicle sequence calculation model is established as follows, with the model decision variable being the attribute state value of the lead vehicle in the sub-queue.

[0074]

[0075]

[0076] Step 3: Check if the output of the model in Step 2 meets the termination condition of the genetic algorithm. If it does, output the optimal virtual lead car sequence; if not, continue to perform selection, crossover and mutation operations in sequence.

[0077] The optimal virtual lead vehicle sequence calculation model is solved using a genetic algorithm. The specific steps of the genetic algorithm in step 3 are as follows:

[0078] Step 3.1: Initialization. The attributes of the first car in the sub-queue within the long queue are used as the solution variables. Individuals are constructed using binary encoding to generate the initial population that meets the requirements. The algorithm parameters are determined as shown in Table 1.

[0079] Table 1. Genetic Algorithm Parameter Table

[0080]

[0081] Step 3.2: Greedy repair. Encode and repair individuals that do not meet the model constraints. If the sub-queue head car attribute is assigned to multiple consecutive cars in the result, the sub-queue vehicle quantity constraint is not valid. Compare the cost values ​​of each car that does not meet the constraint, retain the car with the best cost value among the adjacent cars as the sub-queue head car, and cancel the sub-queue car attribute of the adjacent cars so that the solution result meets the model constraints.

[0082] Step 3.3: Individual assessment: Calculate the fitness of each individual based on the information of all individuals in the population;

[0083] Step 3.4: Determine the generation number. If the generation number has not reached the termination generation number, perform the genetic operations from Step 3.4 to Step 3.6 on the population. If the termination generation number is reached, select the individual with the best fitness performance in the current population as the optimal solution and end the solution for the optimal virtual lead car sequence at this time step.

[0084] Step 3.5: Selection operation. Based on the fitness shown by all individuals in the population, superior individuals are selected through roulette wheel selection and retained for the next generation.

[0085] Step 3.6: Crossover operation. In the current population, chromosomes of two individuals are exchanged in pairs at random crossover locations using a single-point crossover method, resulting in new individuals entering the population.

[0086] Step 3.7: Mutation operation, which alters some genes in all individuals of the population with a mutation probability to produce entirely new individuals. After the genetic operation, the next generation of the population re-enters the evaluation stage of Step 3.2.

[0087] Step 4: The vehicle platoon control model is based on the optimal virtual lead vehicle sequence in Step 3. The speed difference between the lead vehicles of adjacent sub-platoons and the speed difference between the following vehicles and the lead vehicles of the sub-platoons are introduced into the control layer to obtain the sub-platoon lead vehicle control layer model and the sub-platoon follow vehicle control layer model. By synchronously controlling the sub-platoon lead vehicles and the following vehicles, the operating status of all autonomous vehicles in the long platoon is updated and used for the calculation of the optimal virtual lead vehicle sequence in the next time step.

[0088] The vehicle platoon control model is based on the optimal virtual lead vehicle sequence. It divides a long platoon into multiple sub-platoons using virtual lead vehicles. Based on the attributes of the lead vehicle in each sub-platoon of the currently active autonomous vehicle within the long platoon, all following vehicles in the long platoon are distinguished as either lead vehicles or follower vehicles. The decision variables in the long platoon control model are the speed, acceleration, and position of the vehicles specified in the long platoon. The sub-platoon lead vehicle control layer controls all lead vehicles in the sub-platoons within the long platoon, while the sub-platoon follower vehicle control layer controls all follower vehicles in each sub-platoon within the long platoon.

[0089] Currently, platooning vehicles primarily rely on cooperative adaptive cruise control (CACC). The CACC vehicle following model, proposed by the PATH Lab at UC Berkeley, has a simple structure and has been validated through real-vehicle testing. The model parameters are as follows:

[0090]

[0091] In the formula, ei t is the deviation between the actual distance and the desired distance between vehicle i and vehicle i-1, xi t is the displacement of vehicle i at time t, ai t is the acceleration of vehicle i at time t, vi t is the velocity of vehicle i at time t, and T CACC k is the vehicle-to-vehicle time distance, k0 is the acceleration weight of the preceding vehicle in the model, k1 is the vehicle-to-vehicle distance error weight in the model, and k2 is the speed difference weight in the model.

[0092] Based on this queue control, the established vehicle queue control model introduces the speed difference between the lead vehicles of adjacent sub-queues in the long queue and the speed difference between the following vehicles in the sub-queue and the lead vehicle of the sub-queue into the model control layer. Through multiple sub-queues divided by the virtual lead vehicle, effective control of all autonomous following vehicles in the long queue is achieved. After the vehicle state is updated, the optimal virtual lead vehicle sequence is re-inputted into the calculation model. In the loop, the long queue of autonomous vehicles is continuously and stably controlled to ensure the safe driving of vehicles in the long queue.

[0093] According to the optimal virtual lead vehicle sequence calculation model, the autonomous vehicles in a queue with a lead vehicle attribute of 1 possess the lead vehicle attribute. Therefore, the speed difference control parameter between the lead vehicles in the adjacent sub-queues ahead of the lead vehicle can be determined using this attribute.

[0094]

[0095] In the formula, j is the sequential position of the virtual lead car in the long queue, ei t is the deviation between the actual distance and the expected distance between the lead car i and the preceding car i-1 in the sub-queue, ai t+1 is the acceleration of the lead car i in the sub-queue at time t+1, vi t is the velocity of the lead car i in the sub-queue at time t, vi-1t is the velocity of the preceding car i-1 in the sub-queue at time t, vj t is the velocity of the adjacent lead car j in the sub-queue at time t, k0 is the acceleration weight of the preceding car in the sub-queue lead car control model, k1 is the distance error weight in the sub-queue lead car control model, k2 is the speed difference weight of the preceding car in the sub-queue lead car control model, k3 is the speed difference weight between the adjacent lead car in the sub-queue and the long queue lead car in the sub-queue lead car control model, and k4 is the speed difference weight of the adjacent lead car in the sub-queue that does not include the long queue lead car.

[0096] Specifically, the model of the sub-queue lead car control layer is as follows:

[0097]

[0098] In the formula, j is the sequential position of the virtual lead car in the long queue, ei t is the deviation between the actual distance and the expected distance between the lead car i and the preceding car i-1 in the sub-queue, ai t+1 is the acceleration of the lead car i in the sub-queue at time t+1, vi t is the velocity of the lead car i in the sub-queue at time t, vi-1t is the velocity of the preceding car i-1 in the sub-queue at time t, vj t is the velocity of the adjacent lead car j in the sub-queue at time t, k0 is the acceleration weight of the preceding car in the sub-queue lead car control model, k1 is the distance error weight in the sub-queue lead car control model, k2 is the speed difference weight of the preceding car in the sub-queue lead car control model, k3 is the speed difference weight between the adjacent lead car in the sub-queue and the long queue lead car in the sub-queue lead car control model, and k4 is the speed difference weight of the adjacent lead car in the sub-queue that does not include the long queue lead car.

[0099] Based on the output of the optimal virtual lead vehicle sequence calculation model, all following vehicles in the long queue of autonomous vehicles are divided into their respective sub-queues and undergo corresponding queue control under the guidance of the sub-queue leader. The sub-queue follower control layer controls the long queue following vehicles that do not possess the sub-queue leader attribute. Under the constraint of a safe distance, the sub-queue following vehicles are affected by the states of both the vehicle in front of them and the sub-queue leader. The long queue of autonomous vehicles is divided into multiple sub-queues by the virtual lead vehicle. The sub-queue number to which a following vehicle belongs can be determined by the position of the following vehicle and the virtual lead vehicle sequence. Following vehicles belonging to the same sub-queue can obtain the running status information of the sub-queue leader. Based on this, a speed difference control parameter between the sub-queue following vehicles and the sub-queue leader is introduced into the vehicle control. By analyzing the running states of the vehicle in front and the sub-queue leader, the sub-queue follower control layer can determine the acceleration that the vehicle should be given in the next moment, i.e.:

[0100]

[0101] In the formula, ei t is the deviation between the actual distance and the expected distance between the following vehicle i and the preceding vehicle i-1 in the sub-queue, ait+1 is the acceleration of the following vehicle i in the sub-queue at time t+1, vi t is the velocity of the following vehicle i in the sub-queue at time t, vi-1t is the velocity of the preceding vehicle i-1 in the sub-queue at time t, vj t is the velocity of the leading vehicle j in the sub-queue to which the following vehicle belongs at time t, K0 is the acceleration weight of the preceding vehicle in the sub-queue following vehicle control model, K1 is the distance error weight in the sub-queue following vehicle control model, K2 is the speed difference weight of the preceding vehicle in the sub-queue following vehicle control model, and K3 is the speed difference weight of the leading vehicle in the sub-queue to which the following vehicle belongs in the sub-queue following vehicle control model.

[0102] The sub-queue following vehicle control layer distinguishes between two cases based on whether the vehicle belongs to the sub-queue of the lead car in the long queue. When the vehicle belongs to the sub-queue of the lead car in the long queue, the speed difference parameter between the following vehicles and the lead car in the sub-queue is not considered. This is because the control model needs to cover the single-lead car control scenario in the long queue, and the vehicle speed difference parameter has already been considered in the single-lead car control scenario, and there are no other sub-queue lead cars in the long queue. When the vehicle does not belong to the sub-queue of the lead car in the long queue, the speed difference between the following vehicles and the lead car in the sub-queue is added to the vehicle control. Adding the speed difference parameter to the sub-queue can ensure that the following vehicles in the long queue can better match the driving state of the lead car in the long queue.

[0103] Specifically, the model of the sub-queue following vehicle control layer is as follows:

[0104]

[0105] In the formula, ei t is the deviation between the actual distance and the expected distance between the following vehicle i and the preceding vehicle i-1 in the sub-queue, ait+1 is the acceleration of the following vehicle i in the sub-queue at time t+1, vi t is the velocity of the following vehicle i in the sub-queue at time t, vi-1t is the velocity of the preceding vehicle i-1 in the sub-queue at time t, vj t is the velocity of the leading vehicle j in the sub-queue to which the following vehicle belongs at time t, K0 is the acceleration weight of the preceding vehicle in the sub-queue following vehicle control model, K1 is the distance error weight in the sub-queue following vehicle control model, K2 is the speed difference weight of the preceding vehicle in the sub-queue following vehicle control model, and K3 is the speed difference weight of the leading vehicle in the sub-queue to which the following vehicle belongs in the sub-queue following vehicle control model.

[0106] Example 1: Long Queue Control Analysis

[0107] A simulation scenario of the main road of a highway was constructed using SUMO. The optimal control parameter weights for long queue vehicles were determined through long queue control simulation experiments, and the timely response of the virtual lead vehicle sequence calculation model to queue speed disturbances was verified.

[0108] To confirm the control parameter weights of the sub-queue lead car control layer within the long queue control model, based on the existing queue control model, we first analyzed whether adding a speed difference between the lead car and the lead car of a single sub-queue could make the long queue control more stable. The weight of this parameter was increased from 0 to 0.10, an increase of 0.025, and multiple simulation experiments were conducted. The optimal long queue control effect was achieved when the weight of the speed difference parameter between the lead cars of the sub-queues was uniformly set to 0.100.

[0109] To confirm the weight values ​​of the control parameters for the sub-queue following vehicle control layer within the long queue control model, the speed difference parameter between the following vehicle and the lead vehicle in the sub-queue was weighted based on the existing queue control model. The weight value of this parameter was increased from 0 to 0.20, and multiple simulation tests were conducted. Considering both disturbance effects and driving efficiency, the weight value K3 corresponding to the speed difference parameter between the following vehicle and the lead vehicle in the sub-queue within the long queue hierarchical control model was determined to be 0.100.

[0110] The optimal virtual lead vehicle sequence calculation model is obtained by solving a genetic algorithm. The long queue is divided into multiple sub-queues guided by virtual lead vehicles, enabling following vehicles to react promptly to disturbances caused by speed changes in the lead vehicle. During the experiment, the lead vehicle in the long queue undergoes uniform acceleration of 2 m / s² for the first 100 seconds, lasting for 10 time steps, and then uniform deceleration of 3 m / s² for the first 200 seconds, lasting for 20 time steps. The operating status of all autonomous vehicles in the long queue is recorded within the first 2500 steps. The result of the optimal virtual lead vehicle sequence calculation model is output, as shown below. Figure 2 As shown.

[0111] Figure 2 (a) shows the change in the average speed of the long convoy of vehicles in three stages. Figure 2 (b) shows the optimal virtual lead car sequence for each stage under varying queue speeds, with colored blocks representing the lead car of a sub-queue within the longer queue at that moment. For example... Figure 2 As shown, in the first stage, after the long queue enters the control zone, the speed of the following vehicles gradually approaches that of the lead vehicle, and the long queue gradually enters a stable state. Before the 1000th step, the lead vehicle of the sub-queue within the long queue of the autonomous vehicle gradually moves to the latter half of the long queue. The main reason is that the speed of the following vehicles at the front of the long queue has become consistent with that of the lead vehicle, while the following vehicles at the back of the long queue are still adjusting their speed. This proves that the model established in this paper is constantly optimizing the solution based on the current driving state of the vehicles in the long queue. In the second stage, the lead vehicle of the long queue accelerates. At the same moment that the lead vehicle changes speed, the optimal virtual lead vehicle calculation model obtains the optimal solution to adapt to the current speed change, so that the long queue is divided into 8 sub-queues in time, and all the following vehicles in the long queue can respond accordingly to the acceleration. Among them, the vehicles in each sub-queue do not show an even distribution. The sub-queues in the middle of the long queue, such as sub-queue number 6, maintain four following vehicles, indicating that the uniform distribution scheme of the sub-queues does not fully consider the influence of the current driving state of the vehicles. Compared to existing studies that use a uniformly distributed virtual lead car, the virtual lead car sequence calculation model proposed in this paper is based on the vehicle's driving state. It effectively distributes the virtual lead cars at designated positions within the long queue, thus effectively suppressing speed disturbances experienced by following vehicles and achieving better control of the long queue. In the third stage, when the lead car in the long queue decelerates, the model in this paper reacts promptly, dividing the long queue into nine denser sub-queues. This allows all following vehicles in the queue to perceive the speed adjustment of the lead car in the first instance, significantly reducing the adverse effects of speed disturbances.

[0112] Example 2 Model Comparison Analysis

[0113] Based on the SUMO design, a highway simulation environment is created. All vehicles traveling in a long queue on the main highway interact and update their status at constant time steps. Vehicle driving states are input to a hierarchical control model of the long queue via the core control terminal. A genetic algorithm is used to solve for the optimal lead car sequence. The long queue is then divided into multiple sub-queues guided by the determined optimal virtual lead car, and the lead cars and following vehicles in each sub-queue are controlled separately. At a specified time step, the lead car in the long queue undergoes continuous acceleration and deceleration. The speed disturbance caused by the lead car's speed change is transmitted to all following vehicles in the long queue, and all following vehicles respond with corresponding acceleration or deceleration.

[0114] The safety and effectiveness of the model proposed in this paper are verified by comparing the simulation results of the existing PATH laboratory model and the long queue hierarchical control model under accelerated conditions.

[0115] The basic parameters for the simulation experiment are shown in Table 2:

[0116] Table 2. Basic parameters for model comparison simulation experiments.

[0117]

[0118] Depend on Figure 3 It can be seen that the mean and median standard deviation of vehicle speed in the long queue under the hierarchical control model are significantly lower than those of the existing queue control model, indicating that the established model enables the long queue to effectively adapt to speed disturbances, and the queue vehicles continue to drive in a more stable state.

[0119] Energy consumption during the operation of a long convoy is the most intuitive evaluation indicator and needs to be included in the model comparison. Based on various data during the operation of the long convoy, the average energy loss of the vehicles in the convoy within 60s, 120s, and 500s after the lead vehicle accelerates is shown in Table 3. Table 3 shows that within one minute after the lead vehicle accelerates, the average energy consumption of the convoy under the existing model is 1509326J, while the average energy consumption under the model established in this paper is 1360339J, a reduction of 9.87% in average vehicle energy consumption. Within two minutes after the lead vehicle accelerates, the average energy consumption of the convoy under the existing model is 2444988J, while the average energy consumption under the model established in this paper is 2296609J, a reduction of 6.07% in average vehicle energy consumption. Within 500 seconds after the lead vehicle accelerates, the average energy consumption of the convoy under the existing model is 8970809J, while the average energy consumption under the established model is 8797093J, a reduction of 1.94% in average vehicle energy consumption.

[0120] Table 3 Comparison of average energy loss of vehicles in the platoon under the acceleration state of the lead vehicle.

[0121]

[0122] Therefore, the established model makes the following vehicles in a long convoy less affected by the speed disturbance caused by the acceleration of the lead vehicle, and makes the long convoy as a whole exhibit a strong disturbance adaptability. At the same time, the reduction of speed disturbances allows the following vehicles to get rid of the speed oscillation process as soon as possible, effectively reducing the energy loss of vehicles in the long convoy during driving.

[0123] The safety and effectiveness of the model proposed in this paper will be verified by comparing the simulation results of the existing PATH laboratory model and the long queue hierarchical control model under accelerated conditions.

[0124] The basic parameters of the simulation experiment are shown in Table 2.

[0125] Depend on Figure 4 It can be seen that the mean and median standard deviation of vehicle speed in the long queue under the hierarchical control model are significantly lower than those of the existing queue control model, and the standard deviation of speed is always controlled below 1.90 m / s. This indicates that the speed distribution of following vehicles in the long queue is more concentrated under the control of the model established in this paper, and the long queue vehicles handle speed disturbances more reasonably.

[0126] Based on various data collected during the long convoy's movement, the average energy loss of the convoy's vehicles within 60s, 120s, and 500s after the lead vehicle decelerated is shown in Table 4.

[0127] Table 4 Comparison of Average Energy Loss of Vehicles in the Queue Under Lead Vehicle Deceleration Condition

[0128]

[0129] As shown in Table 4, within one minute after the lead vehicle decelerates, the average energy consumption of the platoon under the existing model is 1,338,411 J, while the average energy consumption under the newly established model is 1,164,680 J, representing a 12.98% reduction in average vehicle energy consumption. Within two minutes after the lead vehicle decelerates, the average energy consumption of the platoon under the existing model is 1,898,788 J, while the average energy consumption under the newly established model is 1,722,914 J, representing a 9.26% reduction in average vehicle energy consumption. Within 500 seconds after the lead vehicle decelerates, the average energy consumption of the platoon under the existing model is 5,156,930 J, while the average energy consumption under the newly established model is 4,998,436 J, representing a 3.07% reduction in average vehicle energy consumption. Therefore, the newly established model effectively reduces the speed disturbance caused by the deceleration of the lead vehicle, significantly reducing the speed fluctuation amplitude of following vehicles in a long platoon and minimizing energy loss during the operation of the long platoon.

[0130] The safety and effectiveness of the model proposed in this paper are verified by comparing the simulation results of the existing PATH laboratory model and the long queue hierarchical control model under rapid deceleration conditions.

[0131] The basic parameters of the simulation experiment are shown in Table 2.

[0132] Figure 5 The paper presents a comparison of the speed standard deviations of all vehicles in a long queuing under the condition that the lead vehicle decelerates rapidly. Compared with existing queuing models, the model constructed in this paper has lower mean and median speed standard deviations for vehicles in a long queuing, and the speed standard deviation is always below 3 m / s throughout the process. This indicates that vehicles in a long queuing are less affected by speed disturbances, and the following vehicles have responded reasonably to the speed disturbances caused by the rapid deceleration of the lead vehicle.

[0133] Based on various data during the driving process of the long convoy, the average energy loss of the convoy vehicles within 60s, 120s and 500s after the lead vehicle of the long convoy decelerates rapidly is obtained, as shown in Table 5.

[0134] Table 5 Comparison of average energy loss of vehicles in the platoon under rapid deceleration of the lead vehicle.

[0135]

[0136] As shown in the table, within one minute after the lead vehicle's sudden deceleration, the average energy consumption of the platoon under the existing model was 1,549,520 J, while the average energy consumption under the newly established model was 1,286,106 J, representing a 17.00% reduction in average vehicle energy consumption. Within two minutes after the lead vehicle's sudden deceleration, the average energy consumption of the platoon under the existing model was 1,942,755 J, while the average energy consumption under the newly established model was 1,673,921 J, representing a 13.84% reduction in average vehicle energy consumption. Within 500 seconds after the lead vehicle's sudden deceleration, the average energy consumption of the platoon under the existing model was 3,906,202 J, while the average energy consumption under the newly established model was 3,662,872 J, representing a 6.23% reduction in average vehicle energy consumption. These data demonstrate that the newly established model can significantly reduce the energy consumption of the platoon, especially during periods of vehicle speed oscillation, where the model can reduce energy loss by 17% for each vehicle in a long platoon.

[0137] In summary, compared with existing queuing control models, the proposed model can reasonably respond to the speed fluctuations of the lead vehicle in various acceleration and deceleration scenarios. While ensuring the safety of long queuing, it effectively controls the speed oscillation process of following vehicles, effectively reduces the impact of speed disturbances on following vehicles in long queuing, reduces the average energy consumption of vehicles in long queuing, and improves the stability of long queuing.

[0138] Example 3: Impact Analysis of Long Queues on Vehicle Number

[0139] When a long convoy of autonomous vehicles is traveling on a highway, the number of vehicles in the entire convoy is a core factor in the hierarchical control model.

[0140] During the experiment, the long convoy of autonomous vehicles traveled at a constant speed of 25 m / s. The lead vehicle of the convoy accelerated uniformly at 2 m / s² 5 seconds after the start of the simulation, and continued for 20 time steps. The running status of all autonomous vehicles in the convoy within 4000 steps after the lead vehicle accelerated was recorded. The total number of vehicles in the convoy increased from 10 to 30, an increase of 5. Multiple sets of simulation experiments were conducted, and the basic parameters of the simulation experiments are shown in Table 6.

[0141] Table 6. Experimental Parameters for the Impact of Total Number of Vehicles in the Queue

[0142]

[0143] Based on the simulation data, a graph showing the relationship between the average speed of vehicles in long queues and time under different total numbers of vehicles in long queues is drawn, as shown in Figure 6.

[0144] according to Figure 6 The time it takes for all vehicles in a long queue to accelerate is obtained for different total numbers of vehicles in the queue. In order to evaluate the effect of the vehicle in front of it on the vehicle in the long queue of the autonomous vehicle, the average delay time for the vehicles in the long queue to accelerate is obtained, as shown in Table 7.

[0145] Table 7 Comparison of Acceleration Time for Vehicles in a Queue

[0146]

[0147]

[0148] As shown in Figure 6, when the speed of the lead vehicle in a long queue changes, all following vehicles in the queue are affected by the speed disturbance, and the speed of the vehicles in the queue fluctuates repeatedly within the 500th time step. As the table shows, with the increase in the total number of vehicles in the long queue, the time required for all following vehicles to accelerate gradually increases, but the average delay time for the vehicles to accelerate shows a decreasing trend. This indicates that long queue control improves the driving efficiency of the vehicles in the queue, allowing following vehicles to reach the current speed of the lead vehicle more quickly. On the other hand, Figure 1 Within the first 100 time steps before the lead car in a long convoy accelerates, the fewer vehicles in the convoy, the greater the upward amplitude of the speed oscillation of the following vehicles. However, the speed oscillation also ends earlier. This is mainly because in long convoys with fewer vehicles, fewer following vehicles are affected by the speed oscillation during the speed disturbance transmission process, and the impact of the speed disturbance is concentrated on the last vehicle in the convoy. When the number of vehicles in the convoy is large, the last vehicle is affected by the superimposed impact of the speed disturbance transmission, which in turn restricts the upward fluctuation trend of the last vehicle's speed. In summary, when the total number of vehicles in a long convoy increases, the initial speed oscillation of the following vehicles in the long convoy has a smaller upward amplitude, and the average delay time for the following vehicles in the long convoy to complete acceleration is shorter.

[0149] Example 4: Analysis of the Influence of Acceleration on the Lead Vehicle in a Long Queue

[0150] When a long convoy of autonomous vehicles is traveling on a highway, the acceleration of the lead vehicle is the key factor causing speed fluctuations in all following vehicles in the convoy. Therefore, the acceleration of the lead vehicle in the convoy is used as a variable to analyze the impact of the acceleration of the lead vehicle on the driving of the convoy through simulation results.

[0151] The first part covers the case where the acceleration is positive. The basic parameters for the simulation experiment are shown in Table 8.

[0152] Table 8. Experimental Parameters for Analysis of the Influence of Acceleration on the Lead Vehicle in the Contingent

[0153]

[0154]

[0155] Based on the simulation data, a graph showing the relationship between the standard deviation of the speed of vehicles in a long convoy and time under different acceleration conditions is plotted, such as... Figure 7 As shown. According to Figure 7 The extreme values ​​of the standard deviation of the queue velocity under different acceleration conditions are shown in Table 9. Figure 7 As shown, the acceleration behavior of the lead vehicle in a long convoy causes speed oscillations in the following vehicles. The three curves show a consistent trend in the speed standard deviation. The greater the acceleration of the lead vehicle, the greater the increase in the speed standard deviation of the convoy vehicles.

[0156] Table 9. Extreme values ​​of standard deviation of platoon speed under acceleration conditions of the lead vehicle.

[0157]

[0158] As shown in Table 9, at time step 100, the standard deviation of the vehicle speed in the long platoon reaches its maximum value. Specifically, the maximum standard deviation is 2.58 m / s when the acceleration is 3.0 m / s², and 1.68 m / s when the acceleration is 1.0 m / s², indicating that smaller accelerations cause less speed disturbance. At time step 660, the standard deviation of the vehicle speed in the long platoon reaches its maximum value, indicating that the speed disturbance completely dissipates after the lead vehicle accelerates for 61 seconds, and the long platoon enters a stable acceleration process. At time step 3250, the standard deviation of the vehicle speed in the platoon is below 0.05 m / s. After this point, the long platoon basically catches up with the speed of the lead vehicle, and the long platoon continues to travel at a stable, uniform speed.

[0159] The second part covers the case where the acceleration is negative. The basic parameters of the simulation experiment are shown in Table 10.

[0160] Table 10 Basic Parameters for Experimental Analysis of the Influence of Deceleration on the Lead Vehicle in the Contingent

[0161]

[0162] Based on the simulation data, a graph showing the relationship between the standard deviations of the speeds of vehicles in a long convoy under different acceleration conditions is plotted, such as... Figure 8 As shown.

[0163] according to Figure 8 The extreme values ​​of the standard deviation of the queue under different deceleration conditions are shown in Table 11.

[0164] Table 11 Extreme values ​​of standard deviation of platoon speed under deceleration of the lead vehicle.

[0165]

[0166] Depend on Figure 8 As shown, the deceleration of the lead vehicle in a long convoy causes speed disturbances in the following vehicles with the same trend as those caused by acceleration. As the deceleration of the lead vehicle gradually increases, the speed fluctuations of the convoy vehicles become more pronounced. As shown in Table 11, at time step 100, the standard deviation of the convoy vehicle speed reaches its maximum value. Specifically, the maximum standard deviation is 2.29 m / s when the deceleration is 3.0 m / s², and 1.41 m / s when the deceleration is 1.0 m / s², indicating that the magnitude of the deceleration determines the degree of speed disturbance. At time step 660, the standard deviation of the convoy vehicle speed also reaches a maximum value, indicating that the following vehicles in the convoy enter a stable deceleration process 61 seconds after the lead vehicle decelerates. At time step 3285, the standard deviation of the convoy vehicle speed is below 0.05 m / s, after which the convoy vehicles essentially catch up with the speed of the lead vehicle, and the convoy continues to travel at a constant speed at the current lead vehicle speed. Combining the queue data under the acceleration behavior of the lead vehicle, it can be concluded that the maximum value of the speed standard deviation corresponding to the acceleration behavior of the lead vehicle at a time step of 100 is higher than that of the deceleration behavior of the lead vehicle, while the queue speed standard deviations are basically equal at a time step of 660.

[0167] In summary, when the lead vehicle in a long platoon changes gears, the greater the acceleration, the more severe the speed disturbance. The speed disturbance has the greatest impact 10 seconds after the lead vehicle changes gears, and the standard deviation of the platoon's vehicle speeds stabilizes and decreases 66 seconds after the change, with the speeds of the following vehicles in the long platoon continuously converging towards the speed of the lead vehicle. On the other hand, the speed disturbance caused by the acceleration of the lead vehicle in a long platoon is stronger during speed oscillations than that caused by the deceleration of the lead vehicle, but the speed disturbance caused by the acceleration of the lead vehicle dissipates faster.

[0168] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A hierarchical control model method for long queues under a multi-virtual lead car structure, characterized in that, Includes the following steps: Step 1: The roadside control unit collects driving data of long queues of vehicles; Step 2: Load the driving data described in Step 1 into the optimal virtual lead vehicle sequence calculation model; the optimal virtual lead vehicle sequence calculation model is... In the formula For the first The efficiency cost of a following vehicle. For the first The stability cost of a following vehicle. For the lead car in the long queue t The running speed at any given moment, Let be the acceleration of the lead car in the long convoy at time t. For the first in the long queue A following vehicle t The running speed at any given moment, For the first in the long queue A following vehicle t acceleration at any moment T The safe headway for maintaining the queue is set to 6 time steps, or 0.6 seconds. For the attribute status value of the lead car in the subqueue, when The first time in the queue equal to 1 The following vehicle has the attribute of being the lead vehicle in a sub-platoon, when The 1st time in the queue that is not equal to 1 The following vehicle does not have the attribute of being the lead vehicle in a sub-platoon. N The total number of autonomous vehicles. n This represents the total number of virtual lead vehicles. Step 3: Check if the output of the model in Step 2 meets the termination condition of the genetic algorithm. If it does, output the optimal virtual lead car sequence; otherwise, continue with the selection, crossover, and mutation operations in sequence. The specific steps of the genetic algorithm in Step 3 are as follows: Step 3.1: Initialization, using the attributes of the first car in the sub-queue within the long queue as the solution variable, constructing individuals using binary encoding, generating an initial population that meets the requirements, and determining the algorithm parameters; Step 3.2: Greedy repair. Encode and repair individuals that do not meet the model constraints. If the sub-queue head car attribute is assigned to multiple consecutive cars in the result, the sub-queue vehicle quantity constraint is not valid. Compare the cost values ​​of each car that does not meet the constraint, retain the car with the best cost value among the adjacent cars as the sub-queue head car, and cancel the sub-queue car attribute of the adjacent cars so that the solution result meets the model constraints. Step 3.3: Individual assessment: Calculate the fitness of each individual based on the information of all individuals in the population; Step 3.4: Determine the generation number. If the generation number has not reached the termination generation number, perform the genetic operations from Step 3.4 to Step 3.6 on the population. If the termination generation number is reached, select the individual with the best fitness performance in the current population as the optimal solution and end the solution for the optimal virtual lead car sequence at this time step. Step 3.5: Selection operation. Based on the fitness shown by all individuals in the population, superior individuals are selected through roulette wheel selection and retained for the next generation. Step 3.6: Crossover operation. In the current population, chromosomes of two individuals are exchanged in pairs at random crossover locations using a single-point crossover method, resulting in new individuals entering the population. Step 3.7: Mutation operation, which alters some genes in all individuals of the population with a mutation probability to produce entirely new individuals. After the genetic operation, the next generation of the population re-enters the evaluation stage of Step 3.

2. Step 4: The vehicle platoon control model is based on the optimal virtual lead vehicle sequence in Step 3. The speed difference between the lead vehicles of adjacent sub-platoons and the speed difference between the following vehicles and the lead vehicles of the sub-platoons are introduced into the control layer to obtain the sub-platoon lead vehicle control layer model and the sub-platoon follow vehicle control layer model. By synchronously controlling the sub-platoon lead vehicles and the following vehicles, the operating status of all autonomous vehicles in the long platoon is updated and used for the calculation of the optimal virtual lead vehicle sequence in the next time step.

2. The hierarchical control model method for long queues under a multi-virtual lead car structure according to claim 1, characterized in that, The model of the sub-queue lead car control layer is as follows: In the formula, For the lead car of the sub-queue With the car in front - The deviation between the actual vehicle distance and the expected vehicle distance. For the lead car of the sub-queue exist t Acceleration at time +1 For the lead car of the sub-queue exist t The speed of time, The car in front of the lead car in the sub-queue - 1 in t The speed of time, The first car in the sub-queue adjacent to the first car in the sub-queue exist t The speed of time, The acceleration weights of the lead vehicle in the sub-queue lead vehicle control model. The weight of the vehicle spacing error in the sub-queue lead vehicle control model. The weights for the speed difference between the leading vehicles in the sub-queue lead vehicle control model. In the sub-queue lead car control model, the speed difference weights between the lead cars of adjacent sub-queues and the lead cars of the long queue are used. The weight is the speed difference between the first cars of adjacent sub-queues that does not include the first car of the long queue.

3. The hierarchical control model method for long queues under a multi-virtual lead car structure according to claim 1, characterized in that, The model of the sub-queue following vehicle control layer is as follows: In the formula, Follow the car in the sub-queue With the car in front The deviation between the actual vehicle distance and the expected vehicle distance is between -1 and 1. Follow the car in the sub-queue The acceleration at time t+1 Follow the car in the sub-queue The velocity at time t The car in front of the car following the sub-queue -1 at time t, The car following the car in the sub-queue is the lead car of the sub-queue. The velocity at time t K 0 represents the acceleration weight of the preceding vehicle in the sub-queue following vehicle control model. K 1 represents the vehicle spacing error weight in the sub-queue following vehicle control model. K 2 represents the weight of the speed difference between the preceding vehicle and the vehicle in the sub-queue following vehicle control model. K 3 represents the weight of the speed difference between the lead vehicle and the sub-queue in the sub-queue following vehicle control model.

Images