CN114545928B - Heterogeneous vehicle queue control method based on self-triggering distributed predictive control - Google Patents

Abstract

The invention discloses a heterogeneous vehicle queue control method based on self-triggering distributed predictive control, and belongs to the technical field of intelligent transportation; the method is used for vehicle queue cooperative control, and firstly, a three-order discrete vehicle longitudinal dynamics model is established, and the vehicle queue control problem is equivalently changed into a multi-agent consistency problem through state variable substitution; in order to ensure control performance and avoid communication resource waste as far as possible, a class of state consistency optimization problem of a rolling time domain is designed, and the triggering interval and the control input are optimized simultaneously; in order to improve the online operation efficiency, the analysis relation between the control input and the trigger time is given when offline preparation is performed, the optimization problem only aiming at the trigger time is obtained by bringing the original problem, and the corresponding parameters are obtained and stored; and finally, combining offline data, and carrying out online simulation operation on a row of vehicle queues in different states to obtain the same-speed equidistant vehicle queues, wherein the obtained vehicle queues have the characteristics of good control performance, high operation efficiency and communication resource saving.

Description

Heterogeneous vehicle queue control method based on self-triggering distributed predictive control

Technical Field

The invention relates to the technical field of intelligent transportation, in particular to a heterogeneous vehicle queue control method based on self-triggering distributed predictive control.

Background

In recent years, researches in the related field of intelligent transportation are widely focused, and the problem of vehicle queue control is taken as a part of cooperative application of a vehicle and a road, so that the intelligent transportation system has very important research significance in the aspects of improving the road safety, relieving the traffic pressure, reducing the vehicle oil consumption and the like.

The goal of vehicle fleet control is to keep a line of vehicles traveling on a roadway in an ordered formation according to a prescribed inter-vehicle distance strategy. Most of the vehicle queue control problems at the present stage only consider a time trigger control mode simply, and workshops communicate at fixed time intervals, so that communication resource waste still can be caused. For the deficiency of fixed time triggered communication, an event triggered communication mode is proposed. Conventional event triggering schemes typically design a threshold beyond which triggering communications occur, but still require periodic sampling of state information to determine a threshold condition. The self-triggering mode can obtain the next triggering time through calculation, so that frequent sampling is avoided, and the communication and online calculation efficiency is further improved. The traditional linear control and synovial membrane control methods are difficult to consider the problem of constraint, the existing robust control research is also aimed at a specific formation, and the distributed model prediction control method can display and consider the constraint, is flexibly designed according to a target and has strong anti-interference capability.

In prior art document [1]([1]Zheng Y,Li S E,Li K,et al.Distributed Model Predictive Control for Heterogeneous Vehicle Platoons Under Unidirectional Topologies[J].IEEE Transactions on Control Systems Technology,2017,25(3):899-910.), a distributed control scheme is considered by assigning an optimization problem to each vehicle. The challenges of large-scale calculation optimization are avoided.

In the prior art document [2]([2]Lin.S,Dim.D V.Event-triggered control for vehicle platooning[J].Proceedings oftheAmerican Control Conference,2015,2015:3101-3106.), a method for controlling a fleet in an event triggering manner is proposed, so that the communication efficiency is effectively improved.

Document [3]([3]Li H,Yan W,Shi Y.Triggering and Control Co-design in Self-Triggered Model Predictive Control of Constrained Systems:With Guaranteed Performance[J].IEEE Transactions on Automatic Control, 2018:1-1.) proposes a method for self-triggering and controlling collaborative design, which designs by a distributed model predictive controller, obtains a better control effect and greatly improves communication efficiency. It only proves feasibility for optimization problems and does not give a specific form of analytical solution.

In summary, at present, for a vehicle queue, there is no distributed control method considering a self-triggering mechanism to reduce communication frequency and save on-line optimization time, and at the same time, the optimization problem with control input constraint can be solved.

Disclosure of Invention

The invention aims to provide a novel vehicle queue control method, which predicts the triggering time of future communication by designing a self-triggering mechanism, so as to reduce the sampling times of the vehicle state, reduce the frequency of communication, and simultaneously reduce the calculation times of the optimization problem, thereby reducing the online optimization time. In addition, using distributed model predictive control, optimization problems with control constraints can be displayed in consideration.

In order to achieve the above purpose, the present invention provides the following technical solutions:

A vehicle queue control method based on self-triggering distributed predictive control is characterized in that: comprises the following steps

Step 1: establishing a discrete linear vehicle dynamics state space model:

step 1.1: the longitudinal dynamics of each vehicle consists of an engine, a transmission system, aerodynamic resistance, tire friction, rolling resistance and gravity, and a third-order discrete nonlinear model is as follows

Where i denotes the i-th vehicle, p _i and v _i denote the position and speed of vehicle i, F _i ¹＝(Δtη_tT_i(t))/(m_iR_i) denotes the mechanical transmission force of vehicle i, T denotes time, Δt denotes discrete time intervals, η _t is transmission mechanical efficiency, T _i denotes the actual driving torque of vehicle i, m _i denotes the mass of vehicle i, R _i is the tire radius of vehicle i; Representing the air resistance of vehicle i, C _A is the aerodynamic coefficient; f _i ³＝f_i g Δt represents the friction of the vehicle i, F _i is the rolling resistance coefficient of the vehicle i, g is the inertial acceleration, τ _i represents the inertial hysteresis of the longitudinal dynamics of the vehicle i, the above-mentioned raw model is linearized by accurate feedback, the result is Wherein the method comprises the steps ofRepresenting acceleration of vehicle i, the control input of vehicle i is represented by u _i in the form of

The state of the ith vehicle is defined as x _i＝[p_i,v_i,a_i]^T, and a third-order discrete state space model can be obtained as follows

x_i(t+1)＝A_ix_i(t)+B_iu_i(t) (1.3)

Wherein the method comprises the steps of

Step 1.2: in the invention, the aim of vehicle queue control is that the speeds of all following vehicles can be kept the same finally, and the distance strategy between vehicles is satisfied, and the distance strategy provided by the invention is a fixed distance strategy, and the specific strategy is as follows

Wherein v ₀ is the desired vehicle speed, d ₀ is the desired inter-vehicle distance, and is a constant positive constant. Order theThe new kinetic equation can be obtained as follows:

the vehicle fleet control of (1.3) translates into a consistency problem.

Step 1.3: the communication topological relation of the vehicle queue studied by the invention is shown in a figure I. The overall structure is n vehicles, and the topological relation is any structure based on the PF topological structure. The PF structure is shown in FIG. 2.

Step2: trigger interval and neighbor state design:

step 2.1: defining the triggering time of each vehicle as the triggering interval as To avoid communication failure caused by the gano effect and excessive communication waiting time, an upper and lower boundary is set for the trigger interval, specificallyWherein the method comprises the steps ofIs a natural number. Setting the control time domain to H _i＝sM_i,H_i is a value that varies with M _i.

The control input in the trigger interval is kept unchanged, and is specifically as follows:

wherein, Is shown inAt the moment, for the firstPredicted value of time.

Step 2.2: designing a future state sequence of neighbor vehicles j of each vehicle i

Wherein p.epsilon. { 1.,. The. M _i }, r.e { 0.,. S-1}.

The triggering time of each vehicle i and the neighboring vehicle j is generally different, and the vehicle i receives the triggering time of the neighboring vehicle jIt may happen that the initial trajectory of the vehicle j is identical to the assumed trajectory, but the actual trajectory and the assumed state trajectory are different during the time period of no communication, i.e. during the triggering interval of the vehicle j.

Step3: optimization problem

Step 3.1: and (3) designing an objective function. The overall objective function is divided into two parts, in the form of

Wherein,Referred to as the communication cost function portion. In the communication cost function, the coefficient α _i can be selected, and its size determines whether the objective function is more focused on good control performance or lower communication burden, and the smaller α _i is, the closer to 1, the better control performance is ensured by the system, otherwise, the lower communication frequency is required by the system and the calculation burden is reduced.

The cost function part called cooperative target and control performance is specifically formed as follows

Wherein,In order to be a phase cost function,The term is interpreted as a task in which the state of the vehicle i is as close as possible to the state of all its neighbors j, thereby achieving consistency; An item may be interpreted as achieving the purpose of queue control with as little control input as possible; v _i ^f is a terminal cost function that can be interpreted as the state terminal of vehicle i as small as possible, and that can be interpreted as the ability to maintain the i·d ₀ distance between i and j and other vehicles as close as possible. An appropriate weight matrix Q _ij,R_i,P_ij is selected for each vehicle i while setting the coefficient α _i of the communication cost function.

Step 3.2: assigning an optimization problem to each vehicle

By the steps, preparation is made in advance, and an optimization problem is allocated to each vehicle, specifically as follows

S.t.: formulas (1.6) and

u_{i min}≤u_i≤u_{i max} (1.13)

Wherein rM _i≤q≤(r+1)M_i,p∈{1,...,M_i, r.epsilon.0, S-1, and l.epsilon.0, sM _i -1, the inequality condition for the speed constraint and the state constraint can be obtained from constraint (1.11) as follows

Of the above constraints, (1.11) is an iterative equation constraint condition obtained according to a state equation, and (1.12) is a neighbor state equation constraint condition, which can be brought into an objective function of an optimization problem; (1.13) is a control input constraint and (1.14) is a trigger interval constraint, the set in which contains all natural numbers from minimum interval to maximum interval.

Step 4: offline preparation

Step 4.1: the objective function of the optimization problem is processed first. For any r e {0,., s-1}, p e {1,., M _i, the iterative equation constraint (1.11) can be written as

Wherein,

Step 4.2: for each vehicle, a suitable upper trigger limit is selectedThe lower trigger limit is 1, and each vehicle needs to be solved because each vehicle has a corresponding trigger interval, and each trigger moment, namelyCarrying out calculation, wherein the specific calculation steps are as follows;

step 4.3: the neighbor state equation constraint (1.12) and the iterative equation constraint (1.15) are brought into the objective function (1.10) to obtain

Wherein the method comprises the steps of

Due toIs a constant, so that the optimal control sequence can be obtained by directly solving the objective function of the step.

Step 4.4: the unconstrained optimization problem can easily obtain an analytical solution, but in the invention, the constraint conditions comprise speed constraints and state constraints besides a few of the constraint conditions, namely the optimization problem is an optimization problem with inequality constraints, and the optimization problem is more complex than the unconstrained optimization problem.

Step 4.4.1: considering constraint (1.13), the Lagrangian function is written as

Step 4.4.2: the KKT condition is written in columns if it isIs the optimal value, then can obtain

μ_i≥0,λ_i≥0 (1.23)

Step 4.4.3: when r=s-1, the Lagrangian functions are (1.12), (1.15) and (1.17)

Lagrangian function (1.24) pairConduct derivation and order

Can obtain

Wherein the method comprises the steps of

The method for searching the KKT points comprises the following steps:

(1) Case 1: let λ _i＞0,μ_i =0 at this time I.e. boundary conditions, satisfying the KKT condition, in which case the value is a KKT point;

(2) Case 2: let lambda _i＝0,μ_i > 0 at this time Is also a boundary point, satisfies the KKT condition, and is a KKT point at this time;

(3) Case 3: let lambda _i＞0,μ_i > 0, the optimal value is not constrained at this time, i.e. the constraint condition is not satisfied, and can be regarded as an unconstrained optimization problem, and can be obtained It should be verified whether this point is within the constraint, and if so, the objective function is a convex function, so in this case, this point is the global optimum, both of which can be excluded, otherwise, one of the two points must be the global optimum.

Bringing the three points into (1.24), and comparing to obtainThe smallest point is the global optimal solution.

The global optimum at this time is noted as

Step 4.4.4: when r=s-2, (1.26) and (1.15) are brought into (1.17), and then the

Can obtain

Wherein the method comprises the steps of

Similarly, the KKT point is found and the global optimum is found.

Step 4.4.5: when r=s-3,..1, 0, let

Can be obtained

Wherein the method comprises the steps of

Likewise, search for KKT points, and bring in finding the optimal solution.

Step 4.5: after the work finds the optimal solution, the original optimization problem is brought to calculate each coefficient, and when the calculated optimal solution is within the control constraint range, the optimal solution is equivalent to the unconstrained optimization problem, and the optimal solution can be written into the following form:

Wherein, to

As shown in the above equation, all solutions are brought back into the original objective function, and since the optimization problem at this time has no unknown quantity, each coefficient corresponding to the optimization problem when the i-th vehicle has the optimal solution at the trigger time M _i can be calculated. In the above-mentioned method, the step of,All are coefficients that need to be solved. Otherwise, when the solution falls on the boundary, the corresponding coefficient can be obtained by carrying in.

Step 5: online computing

Step 5.1: initializing, let t=0, and giving a total running time t _run;

Step 5.2: if t is more than t _run, executing step 5.9, otherwise executing step 5.3;

Step 5.3; if it is Sampling and statusTransmitting the vehicle state to the vehicle i, updating the neighbor vehicle state, otherwise, entering step 5.8;

Step 5.4: setting the triggering interval as a constant, solving the original problem to obtain

Step 5.5: will beCarrying out (1.8) to obtain an objective function only comprising a trigger interval M _i;

Step 5.6: according to the stored offline parameters, respectively calculating the objective function value J _i corresponding to each M _i, setting the M _i corresponding to the minimum J _i value as the optimal trigger interval in the form of Will beCarry backObtaining an optimal control input;

Step 5.7: order the

Step 5.8: will beActing on the i car as a control input, and returning to the step 5.2;

step 5.9: and (5) ending the calculation.

Drawings

FIG. 1 is a flow chart of a method of the overall summary of the invention;

FIG. 2 is a communication topology diagram corresponding to the inventive content in the present invention;

FIG. 3 is a diagram of a vehicle communication topology in accordance with an embodiment of the present invention;

FIG. 4 is a vehicle position and speed simulation diagram of an embodiment of the present invention;

Fig. 5 is a simulation diagram of the trigger time according to an embodiment of the present invention.

Detailed Description

The invention will now be further described with reference to examples, figures:

a heterogeneous vehicle queue control method based on self-triggering distributed predictive control specifically comprises the following steps:

Step 1: state equation establishment

The discrete state space model of the present disclosure is as follows

x_i(t+1)＝A_ix_i(t)+B_iu_i(t) (1.31)

A vehicle queue with a sampling interval Δt=0.2 and a number of vehicles of 5, i.e. i e (1, 5), is now selected. Inertial hysteresis τ ₁＝0.51,τ₂＝0.75,τ₃＝0.78,τ₄＝0.70,τ₅ =0.73 of vehicle longitudinal dynamics. The vehicle spacing is selected to be d=20. The initial state of each vehicle is x₁＝[75;15;0],x₂＝[30;17;0],x₃＝[0;20;0],x₄＝[-35;18;0],x₅＝[-85;19;0].

The topology network is a BD structure as shown in fig. 3.

Step 2: trigger interval and neighbor state design

The trigger interval is selected asWherein the method comprises the steps ofS=3, then the control time domain is H _i＝s*M_i =9. Then

To simplify the calculation, the prediction state coefficients of the neighbors are uniformly set to gamma _j＝0.8,j∈N_i.

Step3: optimization problem

Since the cost function isWherein the communication cost is partlyLet α _i =0.8, where i e (1, 5); in the cost function part of the cooperative target and control performance, the coefficient of each punishment term is set as followsR_i＝1，

In the control input constraint, u _imin＝-5,u_imax =5, i.e., -5+.u _i +.5.

Step 4: offline preparation

Step 4.1: for the optimization problem with partial equality constraints, we have obtained before

Wherein the method comprises the steps of

Is provided withThe coefficients of each part are calculated as follows:

Step 4.1.1: when M _i =1

Q_1,1j＝Q_1,5j＝Q_1j,Q_1,2j＝Q_1,3j＝Q_1,4j＝2Q_1j,R_1j＝R_1j＝R_1j＝R_1j＝R_1j＝1,

Q_2,1j＝Q_2,2j＝Q_2,3j＝Q_2,4j＝Q_2,5j＝0,Q_3,1j＝Q_3,5j＝Q_1j,Q_3,2j＝Q_3,3j＝Q_3,4j＝2Q_1j,

Q_4,1j＝Q_4,2j＝Q_4,3j＝Q_4,4j＝Q_4,5j＝0。

Step 4.1.2: when M _i =2

R_1j＝2.63*ones(3),R_2j＝2.822*ones(3),R_3j＝2.78*ones(3),R_4j＝2.9*ones(3), R_5j＝2.428*ones(3);

Step 4.1.3: when M _i =3

R_1j＝4.89*ones(3),R_2j＝5.725*ones(3),R_3j＝5.608*ones(3),R_4j＝5.932*ones(3), R_5j＝4.403*ones(3);

Step 4.2: written Lagrangian function as

When r=2, r=1 and r=0, the parameters in the control input can be found from the above-mentioned found data. The parameters when the trigger interval M _i =1 are now found:

Step 4.2.1: when r=2

Step 4.2.2: when r=1

Step 4.2.3: when r=0

Similarly, when M _i =2 and M _i =3, the above parameters can also be found.

Step 4.3: optimization problem with M _i only:

if the control inputs are all within the inequality constraint, then the result calculated in (2) is brought into the original optimization problem to obtain

Wherein, to

When (when)The various optimization problems can also be found by substituting the results calculated above. The parameters of the new optimization problem are now found when the trigger interval M _i =1.

As can be seen from the foregoing summary of the invention,In practice it is:

Finishing to obtain

Data is carried in to obtain

D_2,1j＝0.001*ones(3),D_2,2j＝0.002*ones(3)，

D_2,3j＝0.002*ones(3),D_2,4j＝0.001*ones(3),D_2,5j＝0.001*ones(3)。

Similarly, when M _i =2 and M _i =3, the above parameters can also be found.

If the control input is partially on the inequality constraint boundary, the parameters to be solved can be obtained by directly carrying the control input.

Step 5: online computing

And storing the offline calculated data in a memory, wherein the specific steps of online calculation are as follows, and the online calculation obtains a simulation result. The total running time is 20 seconds, the simulation diagram of the vehicle position and speed is shown in fig. 4, and the simulation diagram of the vehicle triggering time is shown in fig. 5.

Claims (2)

1. A vehicle queue control method based on self-triggering distributed predictive control is characterized in that: the method comprises the following steps:

step 1: establishing a discrete linear vehicle dynamics state space model:

Step 1.1: a longitudinal vehicle dynamics third-order discrete nonlinear model is established, and the original model is as follows:

Where i denotes the i-th vehicle, p _i and v _i denote the position and speed of vehicle i, F _i ¹＝(Δtη_tT_i(t))/(m_iR_i) denotes the mechanical transmission force of vehicle i, T denotes time, Δt denotes discrete time intervals, η _t is transmission mechanical efficiency, T _i denotes the actual driving torque of vehicle i, m _i denotes the mass of vehicle i, R _i is the tire radius of vehicle i; Representing the air resistance of vehicle i, C _A is the aerodynamic coefficient; f _i ³＝f_i g Δt represents the friction of the vehicle i, F _i is the rolling resistance coefficient of the vehicle i, g is the inertial acceleration, τ _i represents the inertial hysteresis of the longitudinal dynamics of the vehicle i, the above-mentioned raw model is linearized by accurate feedback, the result is Wherein the method comprises the steps ofRepresenting acceleration of vehicle i, the control input of vehicle i is represented by u _i in the form of

The state of the ith vehicle is defined as x _i＝[p_i,v_i,a_i]^T, and the original model (1.1) is written as a three-order discrete state space model:

x_i(t+1)＝A_ix_i(t)+B_iu_i(t)(1.3)

Wherein the method comprises the steps of

Step 1.2: the spacing strategy is given, the specific strategy is as follows

Wherein v ₀ is the expected speed, d ₀ is the expected distance between vehicles, which is a constant normal number, letThe new kinetic equation can be obtained as follows

Step 1.3: constructing a vehicle communication topological relation;

Step2: trigger interval and neighbor state design:

Step 2.1: defining the triggering time of each vehicle i as The triggering interval isSetting an upper and lower boundary for the triggering interval, saidWherein the method comprises the steps ofIs a natural number; setting the control time domain to H _i＝sM_i,H_i is a value that varies with M _i, leaving the control input unchanged during the trigger interval: the control input is in the form of:

wherein, Is shown inAt the moment, for the firstA predicted value of time;

Step 2.2: designing a future error state sequence of a neighbor vehicle j of each vehicle i, wherein the future state sequence is as follows:

Wherein, p is {1,., M _i }, r is {0,., s-1};

step 3: design optimization problem:

step 3.1: the objective function is designed, and the overall objective function is as follows:

wherein, Referred to as a communication cost function portion, in which alpha _i is a coefficient to be determined,The cost function part for cooperative target and control performance is as follows:

wherein, In order to be a phase cost function,As a consistency term for the vehicle i,Is a control item of the vehicle i; The method comprises the steps that a terminal cost function is adopted, and Q _ij,R_i,P_ij is adopted as a weight matrix of the cost function;

Step 3.2: assign optimization problem to vehicle i:

s.t.: formulas (1.6) and

Wherein, rM _i≤q≤(r+1)M_i,p∈{1,...,M_i, r epsilon {0,.. The number, s-1, and l epsilon {0,.. The number, sM _i -1, of the constraints, (1.11) is an iterative equation constraint condition obtained according to a state equation, and (1.12) is a neighbor state equation constraint condition, and all the conditions are brought into an objective function of an optimization problem; (1.13) is a control input constraint and (1.14) is a trigger interval constraint;

step 4: calculating offline parameters:

Step 4.1: processing an objective function of the optimization problem: for any r e {0,., s-1}, p e {1,., M _i }, the iterative equation constraint (1.11) is rewritten as:

wherein,

Step 4.2: for each vehicle, a suitable upper trigger limit is selectedThe lower trigger limit is 1, and each trigger time, namelyCarrying out calculation;

Step 4.3: bringing the neighbor state equation constraint (1.12) and the iterative equation constraint (1.15) into (1.10) yields a new objective function in the form of:

Wherein the method comprises the steps of

Step 4.4: solving the constraint optimization problem by using a KKT method:

step 4.4.1: based on the constraint (1.13), the Lagrangian function is rewritten as:

wherein μ _i,λ_i represents the lagrange multipliers, respectively;

step 4.4.2: the column-written KKT condition is based on the optimal value The method comprises the following steps:

step 4.4.3: solving the problem of optimization when r=s-1, and obtaining Lagrangian functions from (1.12), (1.15) and (1.17) as

Lagrangian function (1.24) pairConduct derivative based on

Obtaining

Wherein the method comprises the steps of

The method for finding the KKT point is as follows:

(1) Case 1: let λ _i＞0,μ_i =0 at this time I.e., boundary conditions, satisfies the KKT condition, where the value is a KKT point,

(2) Case 2: let lambda _i＝0,μ_i > 0 at this timeAnd is also a boundary point, the KKT condition is satisfied, in which case the value is a KKT point,

(3) Case 3: let lambda _i＞0,μ_i > 0, the optimal value is not constrained at this time, i.e. the constraint condition is not satisfied, and can be regarded as an unconstrained optimization problem, and can be obtainedAnd verifying whether the point is within the constraint condition, if so, the point is the global optimum because the objective function is a convex function, both of the above cases can be eliminated, otherwise, one of the two points must be the global optimum,

Bringing three points into (1.24), and comparing to obtainThe smallest point is the global optimal solution, and the global optimal point is

Step 4.4.4: solving the solution of the optimization problem when r=s-2, bringing (1.26) and (1.15) into (1.17), based on

Obtaining

Wherein the method comprises the steps of

Searching a KKT point, and searching a global optimal point;

Step 4.4.5: the r=s-3 is found and, once again, 1,0 solution to the optimization problem:

Obtaining

Wherein the method comprises the steps of

Searching KKT points, and searching an optimal solution;

Step 4.5: after finding the optimal solution, carrying out original optimization problem to obtain each coefficient, and when the obtained optimal solution is in a control constraint range, the optimal solution is equivalent to the unconstrained optimization problem, and can be written into the following form:

Wherein, to

Based on the above equation, all solutions are brought back into the original objective function, and each coefficient of the corresponding optimization problem when the ith vehicle has the optimal solution at the trigger time M _i is calculated, in the above equation,All are coefficients to be solved, otherwise, when solutions fall on the boundary, the corresponding coefficients can be obtained by carrying the solutions in;

step5: solving the vehicle motion state by online operation:

Step 5.1: initializing, let t=0, and giving a total running time t _run;

Step 5.2: if t is more than t _run, executing step 5.9, otherwise executing step 5.3;

Step 5.3; if it is Sampling and statusTransmitting the vehicle state to the vehicle i, updating the neighbor vehicle state, otherwise, entering step 5.8;

Step 5.4: setting the triggering interval as a constant, solving the original problem to obtain

Step 5.5: will beCarrying out (1.8) to obtain an objective function only comprising a trigger interval M _i;

Step 5.6: according to the stored offline parameters, respectively calculating the objective function value J _i corresponding to each M _i, setting the M _i corresponding to the minimum J _i value as the optimal trigger interval in the form of Will beCarry backObtaining an optimal control input;

Step 5.7: order the

Step 5.8: will beActing on the i car as a control input, and returning to the step 5.2;

step 5.9: and (5) ending the calculation.

2. The vehicle queue control method based on the self-triggering distributed predictive control according to claim 1, characterized in that: in step 3.1, the overall objective function (1.8) and its cooperative objective and control performance cost function (1.9) are part.