CN114495578B - Non-signal lamp crossing vehicle scheduling method of multiple virtual fleets based on conflict points - Google Patents

Abstract

The invention discloses a method for dispatching vehicles at intersections without signal lights based on multiple virtual fleets of conflict points, and relates to the field of automatic driving. Based on multiple conflict points at intersections, an overall system composed of multiple virtual fleets is constructed, and vehicles are dispatched at intersections. The conflict relationship is converted into the distance relationship of the virtual fleet, and the throughput of the intersection is converted into the size of the largest circumscribed rectangle in the multi-virtual fleet, and then the conflict-free optimal scheduling problem of the two-dimensional intersection without signal lights is passed through the multi-virtual The fleet is transformed into a one-dimensional bin packing problem. The present invention is oriented to intersections without signal lights, and on the premise of ensuring driving safety, finds an optimal vehicle scheduling strategy for maximizing the throughput of intersections without signal lights.

Description

A Vehicle Scheduling Method for Non-Signaled Intersections Based on Conflict Points with Multiple Virtual Fleets

technical field

The invention relates to the field of linear programming, in particular to a method for dispatching vehicles at intersections without signal lights based on conflict points with multiple virtual fleets.

Background technique

With the development of urbanization and the improvement of people's living standards, the unbalanced growth of the number of vehicles and traffic capacity has brought unprecedented challenges to the transportation system. Fortunately, with continuous breakthroughs in autonomous driving technology in recent years, autonomous driving vehicles have gradually realized complete control of the vehicle and can exchange necessary vehicle information with adjacent vehicles in real time, which has laid a technical foundation for improving the efficiency of the future transportation system. Autonomous driving Vehicles will also become one of the core units for scheduling optimization in the future transportation system.

Traffic intersections, as pre-planned traffic intersection areas, can easily lead to congestion due to their complex topological relationships and changing road conditions. At the same time, nearly 40% of traffic accidents in the United States are related to intersections, so how to achieve The problem of safe and efficient passage at traffic intersections is a common problem faced by countries all over the world. In order to solve this kind of problem, one type of research focuses on the optimal scheduling at traditional signal light intersections. It is mainly divided into two directions. One direction is based on a fixed signal light cycle, scheduling whether the self-driving vehicle passes through the current intersection to meet the remaining time of the current signal light. Maximize the profit below. The other direction is based on dynamic signal lights. The intelligent signal lights at the intersection dynamically adjust the cycle of the signal lights in real time according to the collected vehicle information at the intersection to meet the preset traffic goals.

However, the optimization of traditional signal lights still has unnecessary waste of signal time and unnecessary fuel consumption caused by frequent start and stop before intersections. Therefore, the optimal scheduling of intersections without signal lights has gradually become the core of future traffic scheduling research. There are three main types of scheduling methods. The first type is based on the reservation method. The self-driving vehicle near the intersection will send a traffic request to the centralized solver (controller) device at the intersection. The corresponding space-time resources are reserved for vehicles to pass safely and efficiently. The second type of method is based on the optimization model. The problem of safe and efficient traffic at intersections without signal lights will be abstracted into an accurate optimization model, which will be solved in real time by a centralized high-computing solver placed at the intersection. Vehicle collisions will pass through the model. Several constraints are avoided, and finally different optimization goals are achieved by setting the optimization objective function.

The third type of method is a method based on a single virtual fleet, which is also the most similar realization scheme of the present invention. Different from the previous two methods, this method is a distributed control method, which converts the lane of the two-dimensional non-signal intersection into a one-dimensional virtual lane through the equidistant rotation transformation with the center of the intersection as the fulcrum, and then according to the lane The conflict relationship among them, construct the order of the vehicles entering the intersection in the virtual fleet through the spanning tree method and the first-in-first-out criterion, and finally transform the problem of scheduling safe and efficient traffic of vehicles at two-dimensional intersections into the scheduling of vehicles on one-dimensional virtual lanes. For the control problem of vehicles forming a stable virtual fleet, the distributed control strategy achieves the effect of centralized optimization. Based on the idea of a single virtual fleet, some studies have studied and optimized the sequence of vehicles in a single virtual fleet, striving to achieve higher traffic efficiency. The methods are mainly based on discrete methods such as graph theory, Monte Carlo trees, and network flows. However, since this scheme is a single virtual fleet constructed according to lanes, there is an upper bound on the improvement of intersection utilization, because this scheme only allows vehicles on compatible lanes to drive at the same time, and vehicles in conflicting lanes will be divided into two groups. However, the collision of vehicles will only occur at the intersection of the conflicting lanes, not the entire conflicting lanes. Therefore, there is still room for optimization of the intersection space utilization efficiency of the single virtual fleet scheme.

Therefore, those skilled in the art are devoting themselves to developing a centralized solution and distributed control vehicle scheduling method for intersections without signal lights. The present invention builds a multi-virtual vehicle fleet composed of multiple virtual lanes based on all conflict points in the intersection. Transform the scheduling problem of unsignaled intersections into the multi-virtual convoy packing problem and propose an optimization model based on this problem and its solution method, which effectively solves the problems of difficult control and heavy communication load at unsignaled intersections, and improves the space of intersections Utilization also increases intersection throughput.

Contents of the invention

In view of the above-mentioned defects of the prior art, the technical problem to be solved by the present invention is to find an optimal vehicle scheduling strategy to maximize the throughput of intersections without signal lights under the premise of ensuring driving safety. It mainly solves the problem that the utilization rate of the intersection is not high due to only considering the conflict relationship between lanes; at the same time, it solves the problem of excessive communication and calculation load caused by the frequent calculation of the centralized solver at the intersection, and meets the needs of high traffic flow. Real-time calculation requirements; finally, the present invention only needs the vehicle's lane and relative position to complete the centralized solution of the problem, protecting the privacy data of the vehicle at the intersection.

In order to achieve the above object, the present invention provides a vehicle scheduling method at a non-signalized intersection based on multiple virtual fleets of conflict points, characterized in that the vehicle scheduling method is based on multiple conflict points at the intersection, and a multi-virtual fleet is constructed. The overall system converts the conflict relationship of vehicles at the intersection into the distance relationship of the multi-virtual fleet, converts the throughput of the intersection into the size of the largest circumscribing rectangle in the multi-virtual fleet, and then converts the two-dimensional infinite The conflict-free optimal scheduling problem at signal light intersections is transformed into a one-dimensional box packing problem through the multi-virtual fleet.

Further, the method includes the following steps:

Step 1. Construct multiple virtual lanes to form a virtual fleet;

Step 2, constructing the packing problem, the packing problem under the background of multiple virtual fleets at intersections without signal lights is a packing problem of one-dimensional irregular graphics;

Step 3, optimize the construction of the model;

stp( _xi )≥0

p(x _i )≤d _max

|p(x _j )-p(x _i )|≥d _s , i＜j

Among them, the safe following distance of the vehicle is denoted by d _s , d _max represents the maximum distance between vehicles for intersection throughput or intersection utilization, and p( _xi ) represents the position of vehicle i on multiple virtual lanes;

Step 4, heuristic improvement;

In order to remove some redundant constraints in the optimization model, the present invention proposes corresponding three heuristic improvement strategies to remove redundant constraints:

Strategy 1: The first car entering the coordination area on each actual lane needs to satisfy the first constraint, while the last car needs to satisfy the second constraint. The rest only need to compare whether the relative distance of adjacent vehicles on the same lane meets the safety distance, and there is no need to compare other vehicles in the same lane through the third constraint;

Strategy 2: When the lanes of vehicle i and vehicle j are compatible lanes, there is no need to compare the third constraint;

当某车道上超越车辆i的数量高于阈值

该阈值及之后的车辆都一定不会超越车辆i；具体地，等于阈值

的车辆需满足去掉绝对值后的第三条约束，大于该阈值的车辆无需再与车辆i比较第三条约束；特别地，当

时，表示优化方法基于先入先出的准则；Strategy 3: Predefine an overtaking threshold

When the number of overtaking vehicle i in a lane is higher than the threshold

Vehicles at and after this threshold must never pass vehicle i; specifically, equal to the threshold

Vehicles need to meet the third constraint after removing the absolute value, and vehicles greater than this threshold need not compare the third constraint with vehicle i; especially, when

When , it means that the optimization method is based on the first-in-first-out criterion;

Step 5, form the discrete-time distributed control of the virtual fleet;

When the centralized solver at the intersection obtains the optimal car-following relationship in the steady state of the fleet, the solver will send the information to the vehicles in the coordination area through V2I communication technology, and the vehicles will receive the information that they need to follow in the virtual fleet After the vehicle information and the expected following distance of the vehicle, it will no longer communicate with the centralized solver at the intersection, but instead communicate with adjacent vehicles to obtain the information of the target vehicle in real time and control its own input, so that the vehicle can reach the intersection reach the steady state of the fleet before the zone;

At this time, there are likely to be multiple vehicles that need to follow the vehicle in the virtual fleet, but after strict verification, it is concluded that the vehicle only needs to perform the following process with any one of the vehicles, and the final multi-virtual fleet system will also achieve Desired fleet steady state.

Further, the multi-virtual fleet formation method described in step 1 is that all the vehicles in the lanes passing through the two conflict points determine the absolute position of the vehicle in the virtual lane according to the distance to the conflict point;

The position of the vehicle in the virtual lane is obtained by equidistant transformation of the actual distance between the vehicle and the conflict point, and all the vehicles on each of the virtual lanes must form a stable virtual fleet through the car-following model , so far constructed into the multi-virtual fleet;

可通过如下公式求得：Further, the absolute position of the vehicle in the virtual lane

It can be obtained by the following formula:

为交汇区内车道l_j到冲突点k的裕量距离，该距离可以用离线方法通过几何关系或物理测量得出。where p _i is the distance from vehicle i to the border of the intersection,

is the margin distance from lane l _j in the intersection area to conflict point k, which can be obtained by off-line method through geometric relationship or physical measurement.

Further, the goal of the optimization model in step 3 is to minimize the maximum distance difference between the front and rear vehicles in the virtual fleet, the first two constraints are used to limit all vehicles within the maximum distance difference, and the third constraint It is to ensure that the following distance of any two vehicles in the virtual fleet is greater than the safety distance when the fleet is in a steady state, but this constraint is a non-convex constraint, which is not conducive to the solution of the problem, so a Boolean quantity is introduced to scale the non-convex constraint , and finally convert the optimization model into a form of mixed integer linear programming that is easier for the solver to solve.

Furthermore, the purpose of solving the optimization model is to find the optimal car-following relationship in multiple virtual fleets, and the x _i represented by each vehicle represents the premise of maximizing the utilization of intersection space, when the fleet is stable, the vehicle i is in multiple The relative position in the virtual fleet. At this time, the vehicle in front of vehicle i in any virtual fleet is its optimal car-following object.

Further, the three heuristic improvement strategies in step 4 are based on the following three criteria:

Criterion 1: Two adjacent vehicles on the same lane only need to keep a safe following distance on any of the virtual lanes to ensure that the two vehicles will not collide when the entire virtual fleet passes the intersection; because the adjacent vehicles on the same lane The virtual lanes of the two vehicles are exactly the same, and all vehicles have completed overtaking before entering the coordination area, so the vehicles only need to keep a safe following distance in any virtual lane, which ensures that each other is also in other virtual teams. the same spacing;

Criterion 2: Vehicles on compatible lanes without conflict points will not collide;

Criterion 3: When vehicle j on a certain lane is determined not to overtake vehicle i in the steady state of the fleet, then according to criterion 1, vehicles in the same lane behind vehicle j must not overtake vehicle i.

Furthermore, the centralized solver only needs to solve the optimal vehicle-following relationship once, and the actual control process is realized through the distributed control and communication of the vehicle.

Furthermore, due to the inevitable time interval of V2V communication in the step 5, it is difficult to guarantee continuity in actual vehicle control, so the state space expression of vehicle i in discrete time is as follows:

u_i(k)车辆执行跟车模型的输入量。Where x _i (k)=[p _i (k), v _i (k), u _i (k ^T ) represent the displacement, velocity and acceleration of vehicle i respectively, t _s is the communication time interval of the system, and in the present invention middle

u _i (k) is the input quantity of the vehicle to execute the car-following model.

Further, the input quantity u _i (k) of the vehicle following model: that is, the acceleration of vehicle i is expressed as follows by the linear combination of the speed difference Δv and the expected distance difference Δp with the preceding vehicle j:

u _i (k+1)=k _p Δp+k _v Δv

=k _p (p _j (k)-p _i (k)-D _des )

+k _v (v _j (k)-v _i (k))

Where k _p and k _v are the feedback coefficients of distance difference and speed difference respectively, and the same feedback coefficient is used for all vehicles in the system.

The present invention has the following technical effects:

1. Different from the traditional virtual fleet method based on the lane topological relationship, the present invention constructs multiple virtual fleets based on the conflict points that may collide in the intersection, and maximizes the utilization efficiency of the intersection while ensuring safety. Improved throughput at junctions.

2. The present invention converts the maximum throughput problem at a two-dimensional intersection into a one-dimensional box packing problem on multiple virtual fleets, and uses an optimization model to abstract the optimal scheduling strategy for vehicles at the intersection, and converts it into a solution that is easy for a solver to solve. A form of mixed integer programming.

3. The present invention proposes three heuristic improvement strategies in combination with the actual situation of traffic intersections and driving rules, which removes redundant constraints in the proposed mixed integer programming problem and greatly improves the computational efficiency of the optimization method.

4. The present invention proposes a discrete time vehicle distributed control, which is closer to actual control and easier to implement.

The idea, specific structure and technical effects of the present invention will be further described below in conjunction with the accompanying drawings, so as to fully understand the purpose, features and effects of the present invention.

Description of drawings

Fig. 1 is a crossroad schematic diagram of a preferred embodiment of the present invention;

Fig. 2 is a system structure diagram of a preferred embodiment of the present invention;

Fig. 3 is a preferred embodiment of the present invention to construct multiple virtual lanes;

Fig. 4 is the virtual convoy on the many virtual lanes of a preferred embodiment of the present invention;

Fig. 5 is a caravan steady-state packing problem of a preferred embodiment of the present invention.

Detailed ways

The following describes several preferred embodiments of the present invention with reference to the accompanying drawings, so as to make the technical content clearer and easier to understand. The present invention can be embodied in many different forms of embodiments, and the protection scope of the present invention is not limited to the embodiments mentioned herein.

In the drawings, components with the same structure are denoted by the same numerals, and components with similar structures or functions are denoted by similar numerals. The size and thickness of each component shown in the drawings are shown arbitrarily, and the present invention does not limit the size and thickness of each component. In order to make the illustration clearer, the thickness of parts is appropriately exaggerated in some places in the drawings.

Since there are various types of intersections, the present invention takes the four-way, three-lane intersection shown in FIG. 1 as an example to elaborate on the scheme. There are four entry directions and exit directions at the intersection, and each direction is divided into straight-going, left-turning, and right-turning lanes, and all lanes are named l ₀ -l ₁₁ in a clockwise direction. Two conflicting lanes will meet at a conflict point at the intersection. Since the conflict point of the left-turn lane is very close to the conflict point of the straight lane, several conflict points that are very close to each other are unified into one conflict point, thus forming a For the eight conflict points c ₀ -c ₇ shown in the figure, for the safety of the intersection, multiple vehicles cannot appear at the same conflict point at the same time. In addition, the intersection is also divided into two areas: the intersection area and the coordination area. The former is the intersection area of multiple lanes in the intersection and is also the area most prone to vehicle collisions. The latter is designed for vehicles to dispatch vehicles to safely pass through the intersection. The coordination buffer area, the size of the area is related to the communication range of the intersection facilities.

The system structure of the present invention is shown in Figure 2. All vehicles in the present invention are fully automatic driving vehicles, and are equipped with V2V (Vehicle-to-Vehicle) and V2I (Vehicle-to-Infrastructure) communication without delay or packet loss. technology, and has completed overtaking and lane changing when entering the coordination area. The present invention is based on the vehicle dispatching method of the multi-virtual vehicle fleet of conflict point, and this method mainly comprises the following steps:

Step 1. Build multiple virtual fleets

与

为例，如图3所示，所有经过两个冲突点的车道所在车辆，根据其到冲突点的距离，确定该车辆在虚拟车道中的绝对位置。需要注意的是，本发明中构造虚拟车道的方法不同于传统的单虚拟车道方法，传统方法是将车辆进行等距旋转变换到虚拟车道上，而本发明中车辆在虚拟车道中的位置不是其到对应冲突点的直线距离，而是车辆与冲突点的实际行驶距离通过等距变换得到的。同时，本发明中各虚拟车道的终点是建立该车道所依据的冲突点，而不是传统的路口正中心。Two virtual lanes constructed with conflict points c ₃ c ₄

and

For example, as shown in FIG. 3 , all the vehicles in the lane passing through the two conflict points determine the absolute position of the vehicle in the virtual lane according to the distance to the conflict point. It should be noted that the method of constructing the virtual lane in the present invention is different from the traditional single virtual lane method, the traditional method is to transform the vehicle into the virtual lane by equidistant rotation, but the position of the vehicle in the virtual lane in the present invention is not its The straight-line distance to the corresponding conflict point, but the actual driving distance between the vehicle and the conflict point is obtained through equidistant transformation. At the same time, the end point of each virtual lane in the present invention is the conflict point on which the lane is established, rather than the traditional intersection center.

可通过如下公式求得：After understanding the construction of virtual lanes, each conflict point in the intersection is transformed accordingly, thus obtaining 8 virtual lanes as shown in Figure 4, and all vehicles on each virtual lane must form a car-following model. Stable virtual fleet has constructed 8 virtual fleets so far, i.e. many virtual fleets in the present invention. The absolute position of the vehicle in the virtual lane

It can be obtained by the following formula:

为交汇区内车道l_j到冲突点k的裕量距离，该距离可以用离线方法通过几何关系或物理测量得出。where p _i is the distance from vehicle i to the border of the intersection,

is the margin distance from lane l _j in the intersection area to conflict point k, which can be obtained by off-line method through geometric relationship or physical measurement.

来唯一表示，为了方便起见，本发明使用车辆到其第一个冲突点的虚拟车道上的绝对位置来唯一性的标识该车辆在多虚拟车道上的位置。Since the same lane will pass through different conflict points, the vehicles on the lane will also appear in different virtual lanes. However, since the driving route of the lane is fixed, and the conflict point is also fixed, it is proved through rigorous derivation that the position difference of the same vehicle in different virtual lanes is fixed. In summary, the position of the same vehicle in different virtual lanes in the present invention can be determined by any position

To uniquely indicate that, for the sake of convenience, the present invention uses the absolute position of the vehicle to its first conflict point on the virtual lane to uniquely identify the position of the vehicle on multiple virtual lanes.

Step 2. Construct the bin packing problem

The purpose of constructing multiple virtual lanes is to make the vehicles on each virtual lane form a stable fleet before entering the intersection area, that is, multiple virtual fleets. The same as the formation of the actual fleet, the formation process of the virtual fleet is that the individual vehicles achieve a uniform speed with the entire fleet and a stable expected following distance D _des , as shown in formula (2). We call the final stable multi-virtual fleet state the fleet steady state.

From the perspective of multi-virtual convoys, re-examine intersection conflicts. It can be seen from the steady state of the fleet that all vehicles have a uniform speed at this time, so according to the physical relationship of distance, speed, and time, the time difference between vehicles passing the same conflict point in the steady state of the fleet can be expressed linearly by the distance difference between the front and rear vehicles in the virtual fleet . However, it can be seen from the conflict point at the intersection that there is a time difference between multiple vehicles passing through a certain conflict point, which means that the vehicles will not collide at this conflict point. To generalize the above ideas, if there is a time difference between vehicles passing through each conflict point, then the process of all vehicles passing through the intersection is completely collision-free, that is, passing through the intersection safely. In other words, when every vehicle in the virtual convoy maintains a safe following distance from the vehicle in front, all vehicles will not collide at the intersection.

From the perspective of multi-virtual fleets, re-examine the problem of high-efficiency traffic at intersections. The time for all vehicles to pass through the intersection is the time difference between the first vehicle and the last vehicle, and because the speed of the vehicles in the steady state of the fleet is the same, this time difference can also be simplified as the first vehicle arrives at the first conflict point to The difference between the time when the last vehicle reaches the last conflict point, that is, the difference in distance between the frontmost vehicle and the rearmost vehicle in the virtual convoy. In other words, the smaller the maximum distance between the overall vehicles in the virtual fleet, the higher the utilization rate of the intersection, the shorter the passing time of vehicles, and the higher the throughput of the intersection.

In summary, the present invention represents the car with the required safe following distance of each vehicle as a part of the vehicle, as shown in Figure 5, the safe following distance of the vehicle is represented by d _s , and represents the intersection throughput or The overall maximum vehicle spacing for intersection utilization is denoted by d _max .

So far, the problem of safe and efficient traffic at intersections has been transformed into the problem of finding vehicles with no overlap and small spacing in the virtual fleet. This is similar to the description of the classic two-dimensional box packing problem, that is, arrange several rectangles of different lengths and widths in a fixed-width rectangular container, and minimize the length of the container on the premise of ensuring that all rectangular box packing arrangements do not overlap.

The two-dimensional box packing problem is a classic NP-hard problem, but the problem addressed by the present invention has two particularities. The first point is that the lane where the vehicle is located is fixed, so the width of the rectangle represented by each vehicle is fixed. So the vehicle can only be optimized by moving laterally, not vertically. The second point is that the distance difference between the same vehicle in different virtual lanes is fixed, so the movement of the same vehicle in the container is based on a unified whole composed of multiple identical rectangles as the basic unit of movement, so the smallest unit of scheduling is composed of multiple An irregular figure composed of rectangles. Therefore, the packing problem in the background of multiple virtual fleets at intersections without signal lights is a packing problem of one-dimensional irregular graphics, and the global optimal solution can be obtained by optimizing the method.

Step 3. Construction of optimization model

The optimized model looks like this:

stp( _xi )≥0

p(x _i )≤d _max

|p(x _j )-p(x _i )|≥d _s , i＜j

where p( _xi ) represents the position of vehicle i on multiple virtual lanes. The goal of optimization is to minimize the maximum distance difference between the front and rear vehicles in the virtual fleet. The first two constraints are used to limit all vehicles within the maximum distance difference. The third constraint is to ensure that any two vehicles in the virtual fleet are in a steady state. The following distance of the car is greater than the safety distance, but the constraint is a non-convex constraint, which is not conducive to the solution of the problem. Therefore, the present invention introduces Boolean quantities to scale the non-convex constraint, and finally it is easier to convert the optimization model into a solver Solve the form of the mixed integer linear program.

The purpose of solving the optimization model is to find the optimal car-following relationship in the multi-virtual fleet, and the _xi represented by each vehicle represents the premise of maximizing the utilization of the intersection space, when the fleet is stable, the vehicle i is in the multi-virtual fleet At this time, the front vehicle of vehicle i in any virtual fleet is its optimal car-following object.

It can be seen from the parameters of the optimization model that the solution of the optimization model in the present invention only uses the lane where the vehicle is located and the relative order of entering the intersection coordination area, without using any private information such as the position and speed of the individual vehicle. Therefore, in terms of centralized solution, this scheme can effectively prevent vehicle information from being collected in a large scale and protect the private information of vehicles.

Step 4. Heuristic Improvement

However, when the scale of the above-mentioned model is too large, the phenomenon of constraint explosion will appear, so the time complexity of the solution is very high. In order to remove some redundant constraints in the optimization model, the present invention combines the actual intersection conditions and driving requirements, and summarizes three criteria for vehicle driving at intersections:

Criterion 1: Two adjacent vehicles on the same lane only need to keep a safe following distance on any virtual lane to ensure that the two vehicles will not collide when the entire virtual fleet passes the intersection. Because the virtual lanes of two adjacent vehicles on the same lane are exactly the same, and all vehicles have completed overtaking before entering the coordination area, so the vehicles only need to keep a safe following distance in any virtual lane, which also ensures that they are in the same lane. Same spacing in other virtual convoys.

Criterion 2: Vehicles on compatible lanes without conflict points will not collide.

Criterion 3: When vehicle j on a certain lane is determined not to overtake vehicle i in the steady state of the fleet, then according to criterion 1, vehicles in the same lane behind vehicle j must not overtake vehicle i.

For the above criteria, the present invention proposes three corresponding heuristic improvement strategies to remove redundant constraints:

Strategy 1: The first car entering the coordination area on each actual lane needs to satisfy the first constraint, while the last car needs to satisfy the second constraint. For the rest, it is only necessary to compare whether the relative distance of adjacent vehicles on the same lane meets the safety distance, and there is no need to compare other vehicles on the same lane through the third constraint.

Strategy 2: When the lanes of vehicle i and vehicle j are compatible lanes, there is no need to compare the third constraint.

当某车道上超越车辆i的数量高于阈值

该阈值及之后的车辆都一定不会超越车辆i。具体地，等于阈值

的车辆需满足去掉绝对值后的第三条约束，大于该阈值的车辆无需再与车辆i比较第三条约束。特别地，当

时，表示优化方法基于先入先出的准则。Strategy 3: Predefine an overtaking threshold

When the number of overtaking vehicle i in a lane is higher than the threshold

Vehicles at and after this threshold must never surpass vehicle i. Specifically, equal to the threshold

The vehicle of i needs to satisfy the third constraint after removing the absolute value, and the vehicle greater than this threshold does not need to compare the third constraint with vehicle i. In particular, when

When , it means that the optimization method is based on the first-in-first-out criterion.

Combining the above heuristic improvement strategies can greatly reduce the redundant constraints of the optimization model and save the solution time of the optimization method.

Step 5. Form the discrete-time distributed control of the virtual fleet

When the centralized solver at the intersection obtains the optimal car-following relationship in the steady state of the fleet, the solver will send the information to the vehicles in the coordination area through V2I communication technology, and the vehicles will receive the vehicles that they need to follow in the virtual fleet After information and expected following distance, it will no longer communicate with the centralized solver at the intersection, but instead communicate with adjacent vehicles to obtain the information of the target vehicle in real time and control its own input, so that the vehicle can reach the convoy before reaching the intersection area steady state. It should be noted that at this time, there are likely to be multiple vehicles in the virtual fleet that need to follow the car, but after rigorous verification, it is concluded that the vehicle only needs to follow any one of the vehicles, and the final multi-virtual The fleet system will also achieve the desired steady state of the fleet.

The method of the present invention only needs a centralized solver to solve the optimal car-following relationship once, and the actual control process is realized through the distributed control and communication of the vehicle. This, to a certain extent, reduces the problems of large communication load and high computing power requirements caused by the traditional centralized method due to continuous communication with vehicles.

Due to the inevitable time interval of V2V communication, it is difficult to guarantee continuity in actual vehicle control, so the state space expression of vehicle i in discrete time is as follows:

Where x _i (k)=[p _i (k), v _i (k), u _i (k)T] represent the displacement, velocity and acceleration of vehicle i respectively, t _s is the communication time interval of the system, and in this inventing

The input quantity u _i (k) of the vehicle following model, that is, the acceleration of vehicle i is expressed by the linear combination of the speed difference Δv and the expected distance difference Δp with the preceding vehicle j as follows:

u _i (k+1)=k _p Δp+k _v Δv

＝k _p (p _j (k)-p _i (k)-D _des )+k _v (v _j (k)-v _i (k))

Where k _p and k _v are the feedback coefficients of distance difference and speed difference respectively, and the same feedback coefficient is used for all vehicles in the system.

The preferred specific embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make many modifications and changes according to the concept of the present invention without creative efforts. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art shall be within the scope of protection defined by the claims.

Claims (6)

1. A method for dispatching vehicles at a signal-lamp-free intersection of multiple virtual vehicle fleets based on conflict points is characterized in that an integral system consisting of the multiple virtual vehicle fleets is constructed based on the multiple conflict points of an intersection, the conflict relationship of the vehicles at the intersection is converted into the distance relationship among the multiple virtual vehicle fleets, the throughput of the intersection is converted into the maximum external rectangle size of the multiple virtual vehicle fleets, and then the conflict-free optimal dispatching problem of a two-dimensional signal-lamp-free intersection is converted into a one-dimensional boxing problem through the multiple virtual vehicle fleets; the method comprises the following steps:

step 1, constructing multiple virtual lanes to form a virtual vehicle fleet;

step 2, constructing a boxing problem, namely a boxing problem under the background of multiple virtual fleets at the intersection without the signal lamp, which is a boxing problem of one-dimensional irregular graphs;

step 3, constructing an optimization model;

s.t.p(x _i )≥0

p(x _i )≤d _max

|p(x _j )-p(x _i )|≥d _s ,i＜j

wherein d is used for the safe following distance of the vehicle _s Is represented by d _max Overall vehicle maximum separation, p (x), representing intersection throughput or intersection utilization _i ) Representing the position of the vehicle i on the multiple virtual lanes;

step 4, heuristic improvement;

in order to remove some redundant constraints in the optimization model, three corresponding heuristic improvement strategies are proposed to remove the redundant constraints:

strategy one: the first vehicle entering the coordination area on each actual lane needs to meet a first constraint, the last vehicle needs to meet a second constraint, and the rest vehicles only need to compare whether the relative distance between the adjacent vehicles on the same lane meets the safety distance or not, and do not need to compare other vehicles on the same lane through a third constraint;

and (2) strategy two: when the lanes where the vehicle i and the vehicle j are located are compatible lanes, the third constraint does not need to be compared;

strategy three: specifying a passing threshold

When the number of overtaking vehicles i on a certain lane is higher than the threshold value

The threshold and subsequent vehicles must not overtake vehicle i; in particular, equal to a threshold value

The vehicle(s) need to satisfy a third constraint after the absolute value is removed, and the vehicle(s) larger than the threshold value do not need to compare with the vehicle i for the third constraint; in particular when

Then, the representation optimization method is based on the criterion of first-in first-out;

step 5, forming discrete time distributed control of the virtual motorcade;

when the centralized solver at the intersection obtains the optimal vehicle following relationship when the motorcade is in a steady state, the solver sends information to the vehicles in the coordination area through a V2I communication technology, and after the vehicles receive the information of the vehicles needing to follow in the virtual motorcade and the expected vehicle following distance, the vehicles do not communicate with the centralized solver at the intersection any more, but communicate with the adjacent vehicles to obtain the information of the target vehicle in real time and control the self input so as to ensure that the vehicles reach the steady state of the motorcade before reaching the intersection area;

at the moment, a plurality of vehicles are likely to be required to follow in the virtual fleet, but strict verification shows that the vehicles only need to follow any one of the vehicles to execute the following process, and the final multi-virtual fleet system can also achieve the expected fleet stable state;

the optimization model in the step 3 aims to minimize the maximum distance difference between the front vehicle and the rear vehicle in the virtual fleet, the first two constraints are used for limiting all vehicles within the maximum distance difference, the third constraint is used for ensuring that the following distances of any two vehicles in the virtual fleet are greater than the safe distance when the fleet is in a steady state, but the constraint is a non-convex constraint and is not beneficial to solving the problem, so that Boolean quantity is introduced to zoom the non-convex constraint, and finally the optimization model is converted into a form of mixed integer linear programming which is easier to solve by a solver;

in the step 5, because of the unavoidable time interval of V2V communication, it is difficult to ensure continuity in actual vehicle control, so the state space expression of the vehicle i in discrete time is as follows:

wherein x _i (k)＝[p _i (k),v _i (k),u _i (k)] ^T Respectively representing the displacement, velocity and acceleration of the vehicle i, t _s Is a communication time interval of the system, and

u _i (k) The vehicle executes the input quantity of the following model;

the input quantity u of the vehicle execution following model _i (k) That is, the acceleration of the vehicle i is expressed by a linear combination of the speed difference Δ v and the desired distance difference Δ p with the preceding vehicle j as follows:

u _i (k+1)＝k _p Δp+k _v Δv

＝k _p (p _j (k)-p _i (k)-D _des )+k _v (v _j (k)-v _i (k))

wherein k is _p And k _v Feedback coefficients for the distance difference and speed difference, respectively, and using the same feedback coefficient for all vehicles of the system, D _des Is the desired following distance.

2. The method for dispatching vehicles at the signal-lamp-free intersection of multiple virtual vehicle fleets based on conflict points as claimed in claim 1, wherein the multiple virtual vehicle fleets are formed by determining the absolute position of all the vehicles passing through the two conflict points in the virtual lanes according to the distance from the vehicles to the conflict points in step 1;

the positions of the vehicles in the virtual lanes are obtained by equidistant transformation of the actual driving distances of the vehicles from the conflict points, and all the vehicles on each virtual lane form a stable virtual vehicle fleet through a following model, so far, the virtual vehicle fleet is constructed.

3. The method of claim 2, wherein the absolute position of the vehicle in the virtual lane is determined by the method of dispatching vehicles at the beacon-less intersection of multiple virtual fleets based on the conflict points

Can be obtained by the following formula:

wherein p is _i Is the distance of vehicle i to the intersection junction boundary,

is a lane l in the intersection area _j The distance to the conflict point k, which can be determined off-line by geometric or physical measurements.

4. The method as claimed in claim 3, wherein the purpose of solving the optimization model is to find the optimal following relationship in the multiple virtual vehicle fleets, and each vehicle represents x _i Representation maximizationOn the premise of intersection space utilization rate, the relative position of the vehicle i in the multi-virtual vehicle fleet in the steady state of the vehicle fleet, and at the moment, the front vehicles of the vehicle i in any virtual vehicle fleet are the optimal following objects of the vehicle i.

5. The method for the signal-free intersection vehicle dispatching based on the conflict point multi-virtual vehicle fleet is characterized in that the three heuristic improvement strategies in the step 4 are based on the following three criteria:

criterion one is as follows: two adjacent vehicles on the same lane can ensure that the two vehicles do not collide when the integral virtual vehicle team drives through the intersection only by keeping a safe following distance on any virtual lane; because the virtual lanes where two adjacent vehicles on the same lane are located are completely the same and overtaking is completed before all vehicles enter the coordination area, the vehicles only need to keep a safe following distance on any virtual lane, and the vehicles are ensured to be at the same distance in other virtual fleets;

the second criterion is as follows: the vehicles on the compatible lane without conflict points can not collide;

the third criterion is that: when a vehicle j in a lane has determined that it will not overtake vehicle i at the fleet steady state, then, according to criterion one, a vehicle in the same lane behind vehicle j will never overtake vehicle i.

6. The method as claimed in claim 5, wherein the centralized solver only needs to solve the optimal following relationship once, and the actual control process is realized by distributed control and communication of vehicles.

Images