Multi-objective optimization control method and device for heterogeneous commercial motorcade under time delay condition

Abstract

The invention discloses a heterogeneous commercial motorcade multi-objective optimization control method and device under a time lag condition. The method comprises the following steps: S1, establishing a dynamic model and a queue system model of each vehicle in a commercial motorcade; s2, designing a queue controller based on a distributed model predictive control method, establishing a cost function and a constraint condition, defining a sub-predictive optimization problem for each vehicle, designing a vehicle following error, economical efficiency and comfort cost function, reducing vehicle following errors in heterogeneous vehicle queue driving, reducing energy consumption, improving comfort and improving queue comprehensive performance; s3, a buffer and a delay compensator are integrated on the controller, a delay compensation link is added to a control system, the control effect of the heterogeneous motorcade under the delay condition is improved, and stable control over the heterogeneous motorcade under the non-ideal communication condition is achieved; and S4, the target function is solved by adopting simulated annealing and a particle swarm optimization algorithm, the problem of nonlinear discontinuous target values in the target function is solved according to a probability method, the operation time is shortened, and the solving efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation, and in particular to a multi-objective optimization control method and device for a heterogeneous commercial vehicle fleet under time-delay conditions. Background Art

[0002] Road freight transport has always been the foundation of the world's economic development, and the demand for freight has shown an increasing trend year by year. However, with the substantial increase in road traffic volume, congestion has become increasingly serious, traffic efficiency is low, and the energy consumption of a single vehicle is high. Multi-vehicle platoon control is a typical collaborative motion control. The platoon vehicles can automatically adjust their motion state according to the state information of the surrounding vehicles and achieve the expected platoon geometry through speed control. With the evolution of advanced control theory and the increasing application of fleets in the logistics industry, the multi-objective control problem of platoon operation has become increasingly prominent. The control algorithm should take into account multiple comprehensive performances such as followability, economy, and comfort. In addition, when the number of commercial vehicles increases, the delay will increase and the stability margin of the closed-loop dynamics will decrease. Non-ideal communication with delay has a significant impact on the performance of vehicle platoon control, especially in large-scale platoon systems.

[0003] At present, there are several research methods for commercial vehicle platoon control: PID control, LQR control and distributed model predictive control (DMPC). The advantage of PID control is that the algorithm is simple and relatively mature, but when faced with relatively complex changing systems, the application of PID controllers will lead to unsatisfactory control effects. Although the LQR control algorithm is simple and easy to implement and can obtain the optimal control strategy, it is mainly designed for linear systems and has poor control effects on nonlinear systems. Distributed model predictive control can meet the needs of nonlinear complex systems. For vehicle platoon control, the use of DMPC control algorithm has the advantages of high safety, stable control, and good adaptability to complex working conditions. Therefore, the use of distributed model predictive control for commercial vehicle platoon driving has great research value.

[0004] For the control of heterogeneous platoons, there have been studies on the heterogeneous dynamic characteristics of platoon systems, fixed or variable vehicle spacing strategies, and vehicle-to-vehicle interaction information flow topology. However, most studies have only improved the control effects of heterogeneous platoons, such as stability or economy, without taking into account multiple platoon performances, such as followability, economy, and comfort.

[0005] Communication time delay has a significant impact on vehicle dynamics stability and formation string stability. Common delay processing methods are divided into two types: delay compensation methods based on compensation controllers and information reconstruction. Among them, the compensation controller-based methods mainly include model predictive controllers, H ∞robust controllers and neural network controllers, while the delay compensation method based on information reconstruction mainly solves the communication delay problem. However, most studies focus on homogeneous vehicle platoons with the same system dynamic parameters, and most of them use linear controllers, which have limited application in complex nonlinear systems. Summary of the invention

[0006] In order to solve the problems raised in the above background technology, the purpose of the present invention is to provide a multi-objective optimization control method for heterogeneous commercial vehicle fleets under time-delay conditions, design a distributed model prediction multi-objective controller based on the commercial vehicle queue model, add a delay compensation link to the control system for the time-delay environment, and finally use simulated annealing and particle swarm optimization algorithms to solve the objective function, so as to achieve stable control of the heterogeneous queue followability, economy and comfort under non-ideal communication conditions.

[0007] The present invention adopts the following technical solution:

[0008] The present invention proposes a multi-objective optimization control method for a heterogeneous commercial vehicle fleet under time-delay conditions, including the following:

[0009] S1: Establish the dynamic model of each vehicle in the commercial vehicle platoon and the platoon system model;

[0010] S2: Design a controller for heterogeneous commercial vehicle platoons based on a distributed model predictive control approach, and establish cost functions and constraints;

[0011] S3: Add delay compensation link to the control system;

[0012] S4: Simulated annealing and particle swarm optimization algorithms are used to solve the objective function.

[0013] Furthermore, the dynamic model of each vehicle in S1 is as follows:

[0014] Writing the acceleration as an equation for driving power, braking force, and resistance, according to Newton's second law, the acceleration a i (t), driving force F i,T (t) and resistance F i,r The relationship between (t) can be expressed as

[0015]

[0016] Due to the driving force Formula (1) can be rewritten as

[0017]

[0018] In the formula, m i is the mass of the i-th vehicle, r w,i is the wheel rolling radius of the i-th vehicle, T i(t) is the motor output torque of the i-th vehicle, η m,i is the mechanical efficiency of the transmission system of the i-th vehicle, and the resistance F i,r (t) is composed of rolling resistance, wind resistance and ramp resistance, which can be expressed as

[0019]

[0020] Where g and f are the gravitational acceleration and rolling coefficient respectively, C d,i is the air resistance coefficient of the i-th vehicle, r is the air density, A i The frontal area of ​​the i-th vehicle.

[0021] Furthermore, the queue system model in S1 is as follows:

[0022] Let the system state variables be vehicle displacement, vehicle speed and motor output torque, that is, the state variable x i (t) = [S i (t),v i (t),T q,i (t)] T , the control quantity is the vehicle motor torque, that is, u i (t)T q,i (t),

[0023] Then the state equation is

[0024]

[0025] Rewrite equation (4) as

[0026]

[0027] Where: i is the number corresponding to each vehicle; S i (t) is the displacement of the i-th vehicle; v i (t) is the speed of the i-th vehicle; T i (t) is the actual torque of the motor of the i-th vehicle; e i (t) is the expected torque of the motor of the i-th vehicle; t i is the time-delay parameter of the power system of the i-th vehicle;

[0028] The discretized dynamic equation of the system can be obtained, where △t is the time step of each calculation:

[0029]

[0030] Furthermore, the implementation of S2 includes defining three types of variables, as follows:

[0031] (1) System prediction variables:

[0032] The predicted position of the i-th vehicle at time t+n;

[0033] The predicted speed of the i-th vehicle at time t+n;

[0034] The predicted acceleration of the i-th vehicle at time t+n;

[0035] The predicted control input of the i-th vehicle at time t+n;

[0036] The predicted control output of the i-th vehicle at time t+n;

[0037] (2) Optimal variables:

[0038] The optimal control input for the i-th vehicle at time t+n;

[0039] (3) Assume variables:

[0040] The assumed position of the i-th vehicle at time t+n;

[0041] The assumed speed of the i-th vehicle at time t+n;

[0042] The assumed acceleration of the i-th vehicle at time t+n;

[0043] Hypothetical control input for vehicle i at time t+n;

[0044] Hypothetical control output of vehicle i at time t+n.

[0045] Furthermore, the cost function in S2 includes the following:

[0046] (1) Following car error cost function

[0047] First, establish the following error cost function between the i-th vehicle (i=2,3,4,…,N) and the pilot vehicle; let the system output be The expected state of the pilot car is Then the following error cost function with the pilot car is:

[0048]

[0049] Where: A i is the error weight coefficient matrix between the i-th vehicle and the pilot vehicle;

[0050] Secondly, establish the following error cost function between the i-th vehicle (i=2,3,4,…,N) and the preceding vehicle; define the expected state with the preceding vehicle as Then the following error cost function with the preceding vehicle is:

[0051]

[0052] Where: B i is the error weight coefficient matrix between the i-th vehicle and the preceding vehicle; is the position of the preceding vehicle at time t; is the speed of the preceding vehicle at time t;

[0053] (2) Economic cost function

[0054] J 3,i (k|t)=||C i P i (k|t)·△t||2 (9)

[0056] Where: C i is the energy consumption weight coefficient matrix of the i-th vehicle; P i (k|t) is the motor power of the i-th vehicle;

[0057] (3) Comfort cost function

[0058] By controlling the torque change rate to a smaller level, the comfort of each vehicle in the queue can be ensured. Therefore, the vehicle comfort cost function is:

[0059]

[0060] Where: D i is the comfort weight coefficient matrix of the i-th vehicle; For vehicles in The vehicle torque at a constant speed is as shown in formula (11):

[0061]

[0062] Furthermore, the controller of the heterogeneous commercial vehicle fleet in S2 is as follows:

[0063]

[0064] Where: v min The minimum speed for commercial vehicles on highways; max The maximum speed specified for commercial vehicles on highways; T min is the minimum torque that the vehicle motor can achieve; T max It is the maximum torque that the vehicle motor can achieve.

[0065] Furthermore, the constraint condition of S2 is a penalty function, as follows:

[0066]

[0067] Furthermore, the delay compensation link of S3 includes a buffer, a compensator and a DMPC controller module; the buffer receives data from other vehicles according to the information topology and selects the latest data by comparing with the cached data; the compensator predicts the hypothetical state with delay and obtains the estimated value of the hypothetical state in the prediction time domain; in the DMPC module, a nonlinear longitudinal dynamic model of the vehicle is established, the control input is calculated, and the hypothetical state of the vehicle is sent to the adjacent vehicle; the details are as follows:

[0068] The data transmitted between vehicles include timestamps, vehicle codes, and assumed state trajectories, which are generated by the DMPC controller and represent possible future trajectories based on the optimal solution; assuming that the prediction domain is N p There is a certain delay t when vehicle j transmits information to vehicle i through the communication network ij (t), the hypothetical state trajectory sent by vehicle j at time t and received by vehicle i has a certain time delay compared with the true value, which is defined as:

[0069]

[0070] The buffer and compensator select and predict the state to obtain an estimate that can be used for the DMPC controller calculation; in the buffer, at time t, the newly arrived data With arrival j0 Cache data For comparison, new data is stored and transmitted to the compensator according to the timestamp of the data, while the old data is deleted; when no data arrives at t, the cached data is output

[0071] In the compensator, the error caused by the time lag is compensated to obtain the predicted time domain [t,t+N p -1], that is, the hypothetical state estimate of the leading vehicle and the adjacent vehicle; retain [t,t j -t ij (t j )+N p -1] The assumed state within the prediction range is used, and the prediction algorithm is used to compensate for the excess part. The constant speed method is used to predict the future state of the vehicle. The basic trajectory prediction equation based on the CV method is as follows:

[0072]

[0073] There is a certain error between the estimated state and the actual state of other vehicles. Considering the limitations of vehicle speed, acceleration, torque and prediction time domain, the estimation error is controlled within a certain range. Therefore, there is an upper limit to the estimation error. The hypothetical state estimation at t+k can be expressed as:

[0074]

[0075] In the formula

[0076]

[0077] Then, the estimated hypothetical state trajectory of vehicle j is as follows:

[0078]

[0079] The estimated value Used in DMPC controller to obtain control input with acceptable error.

[0080] Further, the implementation of S4 includes:

[0081] S4.1 Initialize the parameters of each particle; the acceleration value of each vehicle in the fleet constitutes a particle group p1 = {a1, a2, ..., a N ,}, the particle swarm size N is equal to the number of team members, the random update factor c1, annealing rate λ, and maximum number of iterations i are set during initialization max and the minimum fitness value d stop ;

[0082] S4.2 Set the fitness function; take the objective function as one part of the fitness function, and the other part as the constraint set of the penalty function C(k); calculate and record the fitness function value of the initialized particle under the fitness function, d stop Equal to the fitness value, the corresponding particle population is recorded as p g , which is composed of the acceleration value of each vehicle at this time;

[0083] S4.3 Let the initial temperature t0 be the minimum fitness value, that is, t0 = d stop ;

[0084] S4.4 uses random update factor c1, acceleration range and population p g (i) Get a new particle population p g (i+1), then i increases by 1 until the maximum number of iterations i is reached max , as shown in formula (19);

[0085] p g (i+1)=p g (i)+c1[a min ,amax ] N (19)

[0086] S4.5 records the fitness function value d at the current temperature t temp , if d temp ≤d stop , then the accelerated particle population p f Accept and equal d temp The corresponding accelerated particle population p g ; if d temp >d stop , then p f by The probability of accepting p g , so that p f =p g (i+1), or The probability of accepting p g unchanged, as shown in equation (20)-(21);

[0087] if d temp ≤d stop :p f =p g (i+1)(20)

[0088] if d temp >d stop :

[0089]

[0090] S4.6 Return to step S4.4 until i reaches i max , then let i = 1, p g (i) = p f ;

[0091] S4.7 Cooling operation k+1 =λt k ;

[0092] After cooling, S4.8 returns to step S4.4 until the lowest temperature is reached, at which time the acceleration particles and the global optimal value are output.

[0093] The present invention also proposes a multi-objective optimization control device for a heterogeneous commercial vehicle fleet, which can execute the above control method.

[0094] Beneficial effects of the present invention:

[0095] 1. The present invention defines a sub-prediction optimization problem for each vehicle in a heterogeneous fleet, and designs the cost functions of following vehicle error, economy and comfort, so as to reduce the following vehicle error during the driving of a heterogeneous commercial vehicle fleet, reduce energy consumption, improve comfort, and enhance the comprehensive performance of the fleet.

[0096] 2. The present invention adopts simulated annealing and particle swarm optimization algorithm to solve the objective function, and solves the problem of nonlinear discontinuous target value in the objective function according to the probability method, thereby reducing the calculation time and improving the solution efficiency.

[0097] 3. Aiming at the negative impact brought by time lag conditions, the present invention integrates a buffer and a delay compensator on the controller to improve the control effect of heterogeneous commercial vehicle fleets under delay conditions, so as to achieve stable control of heterogeneous queues under non-ideal communication conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 It is a flow chart of the multi-objective optimization control method of a heterogeneous commercial vehicle fleet based on distributed model predictive control in the present invention.

[0099] Figure 2 It is the longitudinal motion force diagram of the commercial vehicle in the present invention.

[0100] Figure 3 It is a flow chart of the simulated annealing and particle swarm optimization algorithm used in solving the objective function in the present invention. DETAILED DESCRIPTION

[0101] This paper proposes a multi-objective optimization control method for heterogeneous commercial vehicle fleets based on distributed model predictive control. Figure 1-3 Further description of the invention:

[0102] Figure 1 This is a flow chart of a multi-objective optimization control method for heterogeneous commercial vehicle fleets based on distributed model predictive control. It includes the following steps:

[0103] S1: Establish the dynamic model of each vehicle in the commercial vehicle platoon and the platoon system model;

[0104] S2: Design a controller for heterogeneous commercial vehicle platoons based on a distributed model predictive control approach, and establish cost functions and constraints;

[0105] S3: Add delay compensation to the control system under non-ideal communication conditions;

[0106] S4: Simulated annealing and particle swarm optimization algorithms are used to solve the objective function.

[0107] Step 1: Establish the dynamic model of each vehicle in the commercial vehicle platoon and the platoon system model. The details are as follows:

[0108] The force of commercial vehicles in the process of moving forward is as follows: Figure 2 As shown, acceleration can be written as an equation for driving power, braking force, and resistance. According to Newton's second law, acceleration a i (t), driving force F i,T (t) and resistance F i,r The relationship between (t) can be expressed as

[0109]

[0110] Due to the driving force Formula (1) can be rewritten as

[0111]

[0112] In the formula, m i is the mass of the i-th vehicle, i g is the speed ratio of the reducer, r w,i is the wheel rolling radius of the i-th vehicle, T i (t) is the motor output torque of the i-th vehicle, η m,i is the mechanical efficiency of the transmission system of the i-th vehicle, and the resistance F i,r (t) is composed of rolling resistance, wind resistance and slope resistance and can be expressed as

[0113]

[0114] Where g and f are the gravitational acceleration and rolling coefficient respectively, and α is the slope of the vehicle. d,i is the air resistance coefficient of the i-th vehicle, r is the air density, Z i The frontal area of ​​the i-th vehicle.

[0115] Let the system state variable x i (t) is the vehicle displacement S i (t), vehicle speed v i (t) and motor output torque T i (t), that is, x i (t) = [S i (t),v i (t),T i (t)] T , control quantity u i (t) is the vehicle motor torque, that is, u i (t) = T i (t), the state equation is

[0116]

[0117] In the formula, N represents the number of vehicles.

[0118] Formula (4) can be rewritten as

[0119]

[0120] Where: i is the number corresponding to each vehicle; S i (t) is the displacement of the i-th vehicle; v i (t) is the speed of the i-th vehicle; T i (t) is the motor output torque of the i-th vehicle; e i (t) is the expected torque of the motor of the i-th vehicle; t i is the time-delay parameter of the power system of the i-th vehicle.

[0121] The discretized dynamic equation can be obtained, where △t is the time step of each calculation.

[0122]

[0123] Step 2: Design the controller of heterogeneous commercial vehicle platoon based on the distributed model predictive control method and establish the cost function and constraints. The details are as follows:

[0124] Based on the modeling of the longitudinal dynamics of commercial vehicle platoons, a sub-predictive optimization problem is defined for each vehicle in the heterogeneous platoon, and the solution is optimized using the information transmitted by the communication topology to obtain the control inputs of the vehicles.

[0125] Let N p For the prediction time domain length of DMPC, the following three types of variables are defined.

[0126] (1) System prediction variables:

[0127] The predicted position of the i-th vehicle at time t+n;

[0128] The predicted speed of the i-th vehicle at time t+n;

[0129] The predicted acceleration of the i-th vehicle at time t+n;

[0130] Predictive control input of the i-th vehicle at time t+n;

[0131] Predicted control output of the i-th vehicle at time t+n.

[0132] (2) Optimal variables:

[0133] The optimal control input for the i-th vehicle at time t+n.

[0134] (3) Assume variables:

[0135] The assumed position of the i-th vehicle at time t+n;

[0136] The assumed speed of the i-th vehicle at time t+n;

[0137] The assumed acceleration of the i-th vehicle at time t+n;

[0138] Hypothetical control input for vehicle i at time t+n;

[0139] Hypothetical control output of vehicle ith at time t+n.

[0140] The cost function for vehicle i in the queue is established below.

[0141] (1) Following car error cost function

[0142] First, establish the following error cost function between the i-th vehicle and the pilot vehicle, i = 2, 3, 4, ..., N. Let the system output be in is the predicted position of the i-th vehicle at time t+k, is the predicted speed of the i-th vehicle at time t+k, is the predicted motor torque of the i-th vehicle at time t+k, and the expected state of the pilot vehicle is in is the assumed position of the pilot vehicle at time t+k, d is the vehicle spacing, is the assumed speed of the pilot car at time t+k, T1 a (k|t) is the assumed motor torque of the pilot vehicle at time t+k, and the following vehicle error cost function is:

[0143]

[0144] Where: A i is the error weight coefficient matrix between the i-th vehicle and the pilot vehicle, and A is taken i =diag(10,...,10).

[0145] Secondly, establish the following error cost function between the i-th vehicle (i=2,3,4,…,N) and the preceding vehicle. Similarly, define the expected state with the preceding vehicle as in is the assumed position of the preceding vehicle at time t+k, d is the vehicle spacing, is the assumed speed of the preceding vehicle at time t+k, is the assumed motor torque of the pilot vehicle at time t+k, then the following vehicle error cost function is:

[0146]

[0147] Where: B i is the error weight coefficient matrix between the i-th vehicle and the preceding vehicle, and B i =diag(5,...,5).

[0148] (2) Economic cost function

[0149] J 3,i (k|t)=||C i P i (k|t)·△t||2 (9)

[0151] Where: C i is the energy consumption weight coefficient matrix of the i-th vehicle, and C i =diag(10,...,10);P i (k|t) is the motor power of the i-th vehicle.

[0152] (3) Comfort cost function

[0153] By controlling the torque change rate to a smaller level, the comfort of each vehicle in the queue can be ensured. Therefore, the vehicle comfort cost function is:

[0154]

[0155] Where: D i is the comfort weight coefficient matrix of the i-th vehicle, and D i =diag(8,...,8); For vehicles in The vehicle torque at a constant speed is shown in formula (11).

[0156]

[0157] From equations (7) to (10), the formation cooperative DMPC controller can be obtained as follows:

[0158]

[0159] In the formula: st represents the constraint condition, J i (t) is the objective function of the i-th vehicle (i=1,2,3,…,N); v min The minimum speed for commercial vehicles on highways; max The maximum speed specified for commercial vehicles on highways; T min is the minimum torque that the vehicle motor can achieve; T max is the maximum torque that the vehicle motor can reach; Np is the predicted time domain length of DMPC.

[0160] The constraint is written as a penalty function C(k):

[0161]

[0162] Where: C i (k) is the penalty function of the i-th vehicle (i=1, 2, 3, …, N).

[0163] Step 3: Under non-ideal communication conditions, add a delay compensation link to the control system.

[0164] The delay compensation link consists of a buffer, a compensator, and a DMPC module. The buffer receives data from other vehicles based on the information topology and selects the latest data by comparing it with the cached data. The delay compensator predicts the hypothetical state with a delay and obtains an estimate of the hypothetical state in the prediction time domain. In the DMPC module, a nonlinear longitudinal dynamic model of the vehicle is established, the control input is calculated, and the hypothetical state of the ego vehicle is sent to the adjacent vehicles. This control structure is applicable to all controlled vehicles in the queue. Next, the buffer and compensator for processing time delays are designed.

[0165] When a non-ideal communication network is used to transmit the status of the preceding and adjacent vehicles, characteristics such as latency and packet loss have a significant impact on the performance of the platoon. When packet loss occurs, the vehicle cannot receive the information. Therefore, it can be considered as an infinite delay, and its unifying feature is the timestamp of the transmitted data. The data transmitted between vehicles includes timestamps, vehicle codes, and hypothetical state trajectories, which are generated by DMPC and represent possible future trajectories based on the optimal solution. Assume that the prediction domain is N p There is a certain delay t when vehicle j transmits information to vehicle i through the communication network ij (t). Vehicle i at time t receives the information of vehicle j at time t ij (t) The hypothetical state sequence w of the vehicle issued before ij a (k|tt ij (t)) is:

[0166]

[0167] Where: is the assumed state of vehicle j, which is defined as vehicle j according to tt ij The optimal control sequence calculated at time (t) is used to control the vehicle and obtain the optimal control sequence in the prediction time domain tt ij (t) Vehicle status at time + k.

[0168] The buffer and compensator select and predict the state to obtain an estimate that can be used for DMPC calculation. In the buffer, at time t, the newly arrived data At the initial time t j0 Cache data arrived The new data is stored and transmitted to the compensator according to the timestamp of the data, while the old data is deleted. When no data arrives at time t, the cached data is output.

[0169] In the compensator, the error caused by the time lag is compensated to obtain the predicted time domain [t,t+N p -1], that is, the hypothetical state estimate of the leading vehicle and the adjacent vehicle. Keep [t,tt ij (t)+N p -1] The hypothetical state within the prediction range is used, and the prediction algorithm is used to compensate for the excess. For the short-term prediction problem of 1s, the prediction method based on kinematics can make accurate predictions at a reasonable calculation speed. In addition, since the accurate dynamic model of the adjacent vehicles cannot be directly obtained, the state prediction using the kinematic model can adapt to most working conditions. Therefore, the constant speed method is used to predict the future state of the vehicle. The basic trajectory prediction equation based on the constant speed method is as follows:

[0170]

[0171] Where: p(t) is the position of the vehicle; △t is the interval time; v(t) is the vehicle speed.

[0172] There is a certain error between the estimated state and the actual state of other vehicles, and the error size is affected by the prediction algorithm. However, considering the limitations of vehicle speed, acceleration, torque and prediction time domain, the estimation error can be controlled within a certain range.

[0173] Therefore, there is an upper limit to the estimation error. The hypothetical state estimate at t+k can be expressed as:

[0174]

[0175] Where: For vehicle j in tt ij (t) The assumed nominal state at time + k; For vehicle j in tt ij (t)+The assumed state at time k; For vehicle j in tt ij The hypothetical state prediction value at time (t)+k is expressed as follows:

[0176]

[0177] Subsequently, vehicle i obtains the estimated value of the hypothetical state sequence as follows:

[0178]

[0179] Where: is the number of vehicles received by vehicle i at vehicle j at tt ij (t) is the estimated value of the hypothetical state sequence of the ego vehicle sent at time (t)+k.

[0180] After the raw data with time delay passes through the buffer, the influence of repeated oscillation of the control input caused by time disorder under time delay can be avoided. The compensator further estimates the current state to reduce the influence of the state error caused by time delay on the controller. In particular, when no new data is received for multiple consecutive samplings, the compensator can obtain a reasonable current state estimate based on the data generation time and the prediction algorithm, instead of using an incorrect constant vehicle state. Subsequently, the estimated value can be It is directly used in the DMPC controller to obtain a control input with acceptable error.

[0181] Step 4: Use simulated annealing and particle swarm optimization algorithm to solve the objective function.

[0182] For the nonlinear discontinuous target value economic index in the objective function, a lookup table is needed to obtain the regenerative braking power under the current speed and braking intensity. The traditional quadratic programming algorithm is difficult to solve. In this case, the heuristic algorithm can solve the problem based on the probabilistic method. The present invention uses simulated annealing and particle swarm optimization algorithms to solve the objective function. The flow chart is as follows: Figure 3 The specific steps are as follows:

[0183] (1) Initialize the parameters of each particle. The acceleration value of each vehicle in the fleet constitutes a particle group p1 = {a1, a2, ..., a N ,}, the particle swarm size N is equal to the number of team members. When initializing, set the random update factor c1, annealing rate λ, and maximum number of iterations r max And the minimum fitness value d stop , where d stop Set it to a larger constant, such as 2000;

[0184] (2) Set the fitness function. The objective function is used as one part of the fitness function; the other part is the constraint set of the penalty function C(k). Calculate and record the fitness function value of the initialized particle under the fitness function. Let d stop Equal to the fitness value, the corresponding particle population is recorded as p g (r), r is the number of iterations, p g (r) is composed of the acceleration value of each vehicle at this time.

[0185] (3) Let the initial temperature be the minimum fitness value, that is, t0 = d stop ;

[0186] (4) Using the random update factor c1, acceleration range and particle population p g (r), and obtain a new particle population p g (r+1), then r is incremented by 1 until the maximum number of iterations r is reached max , as shown in formula (19);

[0187] p g (r+1)=p g (r)+c1[a min ,a max ] N (19)

[0188] (5) Record the fitness function value d at the current temperature T temp , if d temp ≤d stop , then the accelerated particle population p f Accept and equal d temp The corresponding accelerated particle population p g If d temp >d stop , then p f by The probability of accepting p g , so that p f =p g (r+1), or The probability of accepting p g unchanged, as shown in equation (20)-(21);

[0189] if d temp ≤d stop :p f =p g (r+1) (20)

[0191] if d temp >d stop :

[0192]

[0193] (6) Return to step (4) until r reaches r max , then let r = 1, p g (r) = p f ;

[0194] (7) Cooling operation T k+1 =λT k, where λ is the annealing rate, which is 0.96, T k is the temperature after k annealing times;

[0195] (8) After cooling, return to step (4) until the lowest temperature is reached, at which time the acceleration particles and the global optimal value are output.

[0196] Based on the above control method, an embodiment of the present invention further proposes a control device, which may be a central controller of a vehicle or other equipment, and which is capable of executing the above control method.

[0197] The series of detailed descriptions listed above are only specific descriptions of feasible implementation methods of the present invention. They are not intended to limit the scope of protection of the present invention. All equivalent methods or changes that do not deviate from the technical creation of the present invention should be included in the scope of protection of the present invention.

Claims

1. A multi-objective optimization control method for heterogeneous commercial vehicle fleets under time-delay conditions, characterized in that: These include: S1: Establish the dynamic model of each vehicle in the commercial vehicle platoon and the platoon system model; S2: Design a controller for heterogeneous commercial vehicle platoons based on a distributed model predictive control approach, and establish cost functions and constraints; S3: Add delay compensation link to the control system; S4: Simulated annealing and particle swarm optimization algorithms are used to solve the objective function.

2. The multi-objective optimization control method for heterogeneous commercial vehicle fleets under time-delay conditions according to claim 1 is characterized in that: The dynamic model of each vehicle in S1 is as follows: Writing the acceleration as an equation for driving power, braking force, and resistance, according to Newton's second law, the acceleration a i (t), driving force F i,T (t) and resistance F i,r The relationship between (t) can be expressed as Due to the driving force Formula (1) can be rewritten as Where mi is the mass of the i-th vehicle, i g is the speed ratio of the reducer, r w,i is the wheel rolling radius of the i-th vehicle, Ti(t) is the motor output torque of the i-th vehicle, η m,i is the mechanical efficiency of the transmission system of the i-th vehicle. The resistance Fi,r(t) consists of rolling resistance, wind resistance and ramp resistance, which can be expressed as Where g and f are the gravitational acceleration and rolling coefficient respectively, α is the slope of the vehicle, and C d,i is the air resistance coefficient of the i-th vehicle, r is the air density, Z i The frontal area of ​​the i-th vehicle.

3. The multi-objective optimization control method for heterogeneous commercial vehicle fleets under time-delay conditions according to claim 2 is characterized in that: The queue system model in S1 is as follows: Let the system state variables be vehicle displacement, vehicle speed and motor output torque, that is, the state variables The control quantity is the vehicle motor torque, that is, u i (t) = T i (t), Then the state equation is Rewrite equation (4) as Where: i is the number corresponding to each vehicle; S i (t) is the displacement of the i-th vehicle; v i (t) is the speed of the i-th vehicle; T i (t) is the actual torque of the motor of the i-th vehicle; e i (t) is the expected torque of the motor of the i-th vehicle; t i is the time-delay parameter of the power system of the i-th vehicle; The discretized dynamic equation of the system can be obtained, where △t is the time step of each calculation:

4. The multi-objective optimization control method for heterogeneous commercial vehicle fleets under time-delay conditions according to claim 1 is characterized in that: The implementation of S2 includes defining three types of variables, as follows: (1) System prediction variables: The predicted position of the i-th vehicle at time t+n; The predicted speed of the i-th vehicle at time t+n; The predicted acceleration of the i-th vehicle at time t+n; Predictive control input of the i-th vehicle at time t+n; The predicted control output of the i-th vehicle at time t+n; (2) Optimal variables: The optimal control input for the i-th vehicle at time t+n; (3) Assume variables: The assumed position of the i-th vehicle at time t+n; The assumed speed of the i-th vehicle at time t+n; The assumed acceleration of the i-th vehicle at time t+n; Hypothetical control input for vehicle i at time t+n; Hypothetical control output of vehicle i at time t+n.

5. The multi-objective optimization control method for heterogeneous commercial vehicle fleets under time-delay conditions according to claim 4 is characterized in that: The cost function in S2 includes the following: (1) Following car error cost function First, establish the following error cost function between the i-th vehicle and the pilot vehicle, i = 2, 3, 4, ..., N; let the system output be in is the predicted position of the i-th vehicle at time t+k, is the predicted speed of the i-th vehicle at time t+k, is the predicted motor torque of the i-th vehicle at time t+k; and the expected state of the pilot vehicle is in is the assumed position of the pilot vehicle at time t+k, d is the vehicle spacing, is the assumed speed of the pilot car at time t+k, is the assumed motor torque of the pilot car at time t+k, then the following error cost function with the pilot car is: Where: A i is the error weight coefficient matrix between the i-th vehicle and the pilot vehicle, and A is taken i =diag(10,...,10). Secondly, establish the following error cost function between the i-th vehicle (i=2,3,4,…,N) and the preceding vehicle; define the expected state with the preceding vehicle as in is the assumed position of the preceding vehicle at time t+k, d is the vehicle spacing, is the assumed speed of the preceding vehicle at time t+k, is the assumed motor torque of the pilot vehicle at time t+k, and the following error cost function with the leading vehicle is: Where: B i is the error weight coefficient matrix between the i-th vehicle and the preceding vehicle, and B i =diag(5,...,5). (2) Economic cost function J 3,i (k|t)=||C i P i (k|t)·△t||2 (9) Where: C i is the energy consumption weight coefficient matrix of the i-th vehicle, and C i =diag(10,...,10);P i (k|t) is the motor power of the i-th vehicle; (3) Comfort cost function By controlling the torque change rate to a smaller level, the comfort of each vehicle in the queue can be ensured. Therefore, the vehicle comfort cost function is: Where: D i is the comfort weight coefficient matrix of the i-th vehicle, and D i =diag(8,...,8); For vehicles in The vehicle torque at a constant speed is as shown in formula (11):

6. The multi-objective optimization control method for heterogeneous commercial vehicle fleets under time-delay conditions according to claim 5 is characterized in that: The controller of the heterogeneous commercial vehicle platoon in S2 is as follows: Where: J i (t) is the objective function of the i-th vehicle (i=1,2,3,…,N); v min The minimum speed for commercial vehicles on highways; max The maximum speed specified for commercial vehicles on highways; T min is the minimum torque that the vehicle motor can achieve; T max is the maximum torque that the vehicle motor can reach; N p is the predicted time domain length of DMPC.

7. The multi-objective optimization control method for heterogeneous commercial vehicle fleets under time-delay conditions according to claim 6 is characterized in that: The constraint condition of S2 is designed as a penalty function, as follows: Where: C i (k) is the penalty function of the i-th vehicle (i=1, 2, 3, …, N).

8. The multi-objective optimization control method for heterogeneous commercial vehicle fleets under time-delay conditions according to claim 6 is characterized in that: The delay compensation link of S3 includes a buffer, a compensator and a DMPC controller module; the buffer receives data from other vehicles according to the information topology and selects the latest data by comparing with the cached data; the compensator predicts the hypothetical state with delay and obtains the estimated value of the hypothetical state in the prediction time domain; in the DMPC module, a nonlinear longitudinal dynamic model of the vehicle is established, the control input is calculated, and the hypothetical state of the vehicle is sent to the adjacent vehicle; the details are as follows: The data transmitted between vehicles include timestamps, vehicle codes, and assumed state trajectories, which are generated by the DMPC controller and represent possible future trajectories based on the optimal solution; assuming that the prediction domain is N p There is a certain delay t when vehicle j transmits information to vehicle i through the communication network ij (t), vehicle i at time t receives the information of vehicle j at time t ij (t) The hypothetical state sequence w of the vehicle issued before ij a (k|tt ij (t)) is: Where: is the assumed state of vehicle j, which is defined as if vehicle j is ij The optimal control sequence calculated at time (t) is used to control the vehicle and obtain the optimal control sequence in the prediction time domain tt ij (t) Vehicle status at time + k. The buffer and compensator select and predict the state to obtain an estimate that can be used for the DMPC controller calculation; in the buffer, at time t, the newly arrived data At the initial time t j0 Cache data arrived The new data is stored and transmitted to the compensator according to the timestamp of the data, while the old data is deleted; when no data arrives at time t, the cached data is output In the compensator, the error caused by the time lag is compensated to obtain the predicted time domain [t,t+N p -1], that is, the hypothetical state estimate of the leading vehicle and the adjacent vehicle; retain [t, tt ij (t)+N p -1] The assumed state within the prediction range is used, and the prediction algorithm is used to compensate for the excess part. The constant speed method is used to predict the future state of the vehicle. The basic trajectory prediction equation based on the constant speed method is as follows: Where: p(t) is the position of the vehicle; v(t) is the speed of the vehicle. There is a certain error between the estimated state and the actual state of other vehicles. Considering the limitations of vehicle speed, acceleration, torque and prediction time domain, the estimation error is controlled within a certain range. Therefore, there is an upper limit to the estimation error. The hypothetical state estimation at t+k can be expressed as: Where: For vehicle j in tt ij (t) The assumed nominal state at time + k; For vehicle j in tt ij (t)+The assumed state at time k; For vehicle j in tt ij The hypothetical state prediction value at time (t)+k is expressed as follows: Subsequently, vehicle i obtains the estimated value of the hypothetical state sequence as follows: Where: is the number of vehicles received by vehicle i at vehicle j at tt ij (t) is the estimated value of the hypothetical state sequence of the ego vehicle sent at time (t)+k. The estimated value Used in DMPC controller to obtain control input with acceptable error.

9. The multi-objective optimization control method for heterogeneous commercial vehicle fleets under time-delay conditions according to claim 1 is characterized in that: The implementation of S4 includes: S4.1 Initialize the parameters of each particle; the acceleration value of each vehicle in the fleet constitutes a particle group p1 = {a1, a2, ..., a N ,}, the particle swarm size N is equal to the number of team members, the random update factor c1, annealing rate λ, and maximum number of iterations r are set during initialization max and the minimum fitness value d stop , such as 2000; S4.2 Set the fitness function; take the objective function as one part of the fitness function, and the other part as the constraint set of the penalty function C(k); calculate and record the fitness function value of the initialized particle under the fitness function, d stop Equal to the fitness value, the corresponding particle population is recorded as p g (r), r is the number of iterations, p g (r) is composed of the acceleration value of each vehicle at this time; S4.3 Let the initial temperature t0 be the minimum fitness value, that is, t0 = d stop ; S4.4 uses random update factor c1, acceleration range and particle population p g (r), and obtain a new particle population p g (r+1), then r is incremented by 1 until the maximum number of iterations r is reached max , as shown in formula (19); p g (i+1)=p g (i)+c1[a min ,and max ] N (19) S4.5 records the fitness function value d at the current temperature T temp , if d temp ≤d stop , then the accelerated particle population p f Accept and equal d temp The corresponding accelerated particle population p g ; if d temp >d stop , then p f by The probability of accepting p g , so that pf = pg(r+1), or The probability of accepting p g unchanged, as shown in equation (20)-(21); if d temp ≤d stop :p f =p g (r+1) (20) if d temp >d stop : S4.6 Return to step S4.4 until r reaches r max , then let r = 1, p g (r) = p f ; S4.7 Cooling operation T k+1 =λT k , where λ is the annealing rate, T k is the temperature after k annealing times; After cooling, S4.8 returns to step S4.4 until the lowest temperature is reached, at which time the acceleration particles and the global optimal value are output.

10. A multi-objective optimization control device for a heterogeneous commercial vehicle fleet, characterized in that: The control device can execute the control method described in claim 1.

Images