Heterogeneous automatic driving vehicle formation environment-friendly driving method

Abstract

The invention provides a heterogeneous automatic driving vehicle formation environment-friendly driving method, and relates to the technical field of fleet control, and the method comprises the steps: building a vehicle formation longitudinal dynamics system, a fuel consumption model and a PLF communication topological structure based on a heterogeneous vehicle formation of a front vehicle-navigator following type PLF communication topology; establishing a distributed model prediction control algorithm of the heterogeneous vehicle formation; the algorithm specifically comprises an optimal fuel consumption driving control algorithm for a pilot vehicle and a distributed control algorithm for each following vehicle in a motorcade. Setting constraint conditions for the objective function of the open-loop optimal control problem; an OSQP mathematical solver is adopted to solve a fuel optimization problem for a continuous quadratic programming algorithm SQP, namely, an optimal fuel consumption driving control algorithm, the optimal driving speed and driving acceleration of the pilot vehicle are obtained, a predicted value input by control is applied to update the vehicle state at the (k + 1) th moment, and the following vehicle reaches the optimal fuel consumption according to the driving of the pilot vehicle, so that the optimal driving speed and driving acceleration of the pilot vehicle are achieved. And the purposes of optimal fuel consumption and environmental protection are achieved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle fleet control technology, and in particular to an environmentally friendly driving method for a heterogeneous autonomous driving vehicle formation. Background Art

[0002] In recent years, connected-automated-vehicles (CAVs) have attracted attention from industry and academia for their ability to improve safety and traffic efficiency. The goal of CAVs is to form a platoon of vehicles that follows an ideal speed and maintains a safe driving distance. Improving following ability, safety performance, fuel economy, and string stability can be achieved simultaneously. The earliest vehicle platooning experiments can be traced back to the PATH program (Partners for Advanced Transit and Highways) in the 1980s. CAV platooning control is a multi-objective optimization problem for distributed systems. Studies have shown that the tracking, safety, energy economy, and string stability objectives of vehicle platooning are conflicting. In addition, in order to ensure the safety of vehicle platooning, the physical limitations of the vehicles must be considered when designing the platooning controller. Distributed model predictive control (DMPC) can explicitly handle constraints and multiple objective optimization (MO) problems within the framework of optimal control, so the MO problem is a feasible solution to the platooning control problem. In order to balance the conflicts among the objectives of multiple vehicles in a platoon, a common approach in DMPC is to sum up multiple objectives into a single objective and set weights to reflect the relative importance of the objectives. Although the weighted summation method is intuitive and widely used in platoon control, there is no general rule for selecting weights for constrained nonlinear systems. In real traffic scenarios, CAVs in the same platoon are usually heterogeneous vehicles with different weights or powers, and the economic optimal speeds of different vehicles are also different.

[0003] Efforts have been made in vehicle energy conservation control, i.e., each vehicle in a group of CAVs should try to save its own fuel consumption as much as possible through eco-driving control (e.g., economical MPC). This will lead to non-cooperative self-interested driving behavior, which may deteriorate the cooperative control objectives of heterogeneous vehicle platoons, such as tracking and chord stability. Since the energy efficiency function of the vehicle is not positive finite or convex at the economic optimal speed, the vehicle closed-loop system with eco-driving control may lose stability. Therefore, it is necessary to find a distributed eco-driving control method that can take into account both the self-interest of energy conservation and the cooperative control objectives of heterogeneous vehicle platoons, and achieve the stability of the resulting vehicle closed-loop system. Chord stability is another important stability issue of vehicle platooning because it largely determines the impact of CAVs on traffic flow stability. A vehicle platoon is said to be chord-stable if the fluctuations in the cruising state of the vehicle do not propagate along the vehicle chord. Compared with single-vehicle stability, the concept of chord stability is more rigorous, and a vehicle platoon may be unstable even if each vehicle in the platoon has stability.

[0004] However, existing studies are only based on the predecessor-following (PF) communication topology, which has a simple structure but can only obtain the status information of the adjacent preceding vehicle. Han Hailan et al. proved that the stability margin, interference suppression and scalability of adjacent vehicles in a vehicle formation are closely related to the communication topology, and the effect is better as the amount of status information obtained increases. Under the same communication conditions, the predecessor-leader following (PLF) topology can not only obtain the information of the adjacent preceding vehicle, but also the status information of the leader vehicle. Based on the simplification of the vehicle formation communication topology in the above research, the optimization target only focuses on the drastic acceleration changes caused by the minimum following distance, which has problems such as interference with the stability of the vehicle formation string and increased fuel consumption. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an environmentally friendly driving method for a heterogeneous autonomous driving vehicle formation, and proposes a fleet control algorithm that combines the fuel consumption optimization driving strategy of the pilot vehicle with the distributed model predictive control of the vehicle formation.

[0006] A method for environmentally friendly driving of a heterogeneous autonomous driving vehicle formation, comprising the following steps:

[0007] Step 1: Based on the heterogeneous vehicle platooning of the leading vehicle-leader following PLF communication topology, the longitudinal dynamics system of the vehicle platooning, the fuel consumption model and the PLF communication topology are established;

[0008] The heterogeneous vehicle formation consists of a manually driven vehicle as an interference vehicle, a pilot vehicle and a plurality of follower vehicles;

[0009] The vehicle formation longitudinal dynamics system is specifically:

[0010] At time t, the nonlinear model of the vehicle is defined as a one-dimensional coordinate along the lane direction to describe the longitudinal motion of the vehicle, and the starting point of the lane is defined as the origin of the coordinates; first, the driving characteristics of the vehicle under the mutual influence of the front and rear vehicles are described as follows: Let d i represents the following distance, v i (t) and a i (t) represent velocity and acceleration respectively, ε i represents the engine input, represents the speed difference between the i-1th vehicle and the i-th vehicle, It means that the acceleration is obtained by taking the derivative of the velocity of the i-th vehicle. Represents the acceleration a i (t) Derivative to get the jerk The degree of change of reaction acceleration, i is the vehicle number:

[0011]

[0012] f in formula (1) i (v i ,a i ) and g i (v i ,a i ) are:

[0013]

[0014] Improve the nonlinear model to describe the motion characteristics of each vehicle in platooning affected by air resistance;

[0015]

[0016] where τ i is the dynamic delay coefficient, σ is the air resistance coefficient, m i , are the mass, cross-sectional area, mechanical resistance and drag coefficient of vehicle i, A i is the air resistance matrix; the engine input ε i Expressed as

[0017]

[0018] Among them, u i Represents the control input; converts the nonlinear system into a linear system acceleration derivative Expressed as

[0019]

[0020] Assume that the state of each car is The output is represented as Formula (1) is expressed as

[0021]

[0022] The input matrix Output Matrix State Matrix Where Δt represents the sampling interval;

[0023] Define X(t), Y(t) and U(t) as the state vector, output vector and input vector of each vehicle, namely:

[0024]

[0025] The state vector from the first car to the Nth car is expressed as

[0026] The output vector from the 1st car to the Nth car is expressed as

[0027] The input vector from the 1st car to the Nth car is expressed as

[0028] The dynamic state equation of the vehicle formation is expressed as:

[0029]

[0030] in B′=diag[B1,...,B N ],

[0031] A′, B′, and C′ represent the vehicle’s state matrix, input matrix, and output matrix, respectively; A N represents the state matrix of the Nth vehicle, B N represents the input matrix of the Nth vehicle, Represents the N-order unit matrix and output matrix The Kronecker product of

[0032] The control goal of the vehicle formation is to maintain a zero speed difference and a desired safe distance with the vehicle in front, that is,

[0033]

[0034] In summary, the longitudinal dynamic system of the vehicle formation is obtained as shown in equation (6);

[0035] The fuel consumption model is specifically:

[0036] Use the VT-CPFM model to estimate the vehicle's real-time fuel consumption:

[0037]

[0038] Among them, F i (t) represents the instantaneous fuel consumption of vehicle i, P i (t) is the real-time power of the vehicle, β0, β1, β2 are model parameters; the vehicle power is related to acceleration and speed, satisfying:

[0039]

[0040] Among them, R i is the vehicle resistance, ρ is the air density, λ i is the rolling friction coefficient, θ(t) is the road slope, η is the vehicle transmission efficiency, is the air resistance coefficient

[0041]

[0042] The PLF communication topology structure is specifically:

[0043] Using directed graphs represents the communication topology of the fleet, is the node set, N is the total number of nodes, It is the communication link between neighboring cars. Contains three attribute matrices: adjacency matrix Laplacian Matrix and traction matrix The adjacency matrix It is used to represent the directional communication between neighboring vehicles and is defined as a ij It is an element in the real number matrix, using 0 or 1 to indicate whether there is a communication link between vehicles, 0 means no link, and 1 means there is a link; Represents a real matrix of order N, namely

[0044]

[0045] in Indicates that there is a directed edge between node j and node i, that is, vehicle i can receive the status information of vehicle j; Laplace matrix Defined as in, is the in-degree matrix, defined as in, Indicates that node i is in-degree; traction matrix Indicates whether there is a communication connection between the following vehicle and the leading vehicle; Wakabe Then p i =1, indicating that there is directed communication between the following vehicle i and the leading vehicle, otherwise, p i = 0; the set indicating whether the pilot vehicle information is transmitted to vehicle i is expressed as

[0046]

[0047] Step 2: Establish a distributed model predictive control algorithm for heterogeneous vehicle platoons; specifically, it includes an optimal fuel consumption driving control algorithm for the lead vehicle and a distributed control algorithm for each following vehicle in the platoon.

[0048] The optimal fuel consumption driving control algorithm for the pilot vehicle is specifically as follows:

[0049] The nonlinear model predictive control is combined with the fuel consumption model. Under physical constraints, the speed information of the preceding vehicle is used to calculate the control input u0 with the best fuel economy. At the same time, the distance constraint d0 and the vehicle speed constraint v0 are used to constrain the vehicle acceleration a0 and the control input u0:

[0050]

[0051] Where J0 is the prediction interval T P Fuel consumption within the range; at time k, the predicted state of the pilot car is s, which is the prediction interval range, s∈[0,T P ]; It is about the predicted speed and acceleration at each time step at time k d 0max d 0min are the maximum and minimum following distance limits, d 0，k+s is the following distance of the sth prediction interval at time k; v 0max 、v 0min are the maximum and minimum following speed limits, v 0，k+s is the following vehicle speed in the sth prediction interval at time k; a 0max 、a 0min are the maximum and minimum following acceleration, a 0，k+s is the following vehicle acceleration in the sth prediction interval at time k;

[0052] The distributed control algorithm of each following vehicle in the convoy is:

[0053] Establish the objective function of the open-loop optimal control problem: the following vehicle obtains the information of the preceding vehicle through the V2V communication device and solves the optimal control problem; assuming that there is no communication delay, the same length of prediction range N is used for the output and input vectors of each vehicle p, in the prediction range [t,t+N p ], the following trajectories are defined: represents the output trajectory predicted by the local optimal control problem; It represents the optimal solution to a local problem solved numerically; represents the assumed output trajectory, Indicates that the output trajectory is assumed to be transferred in reverse; define the control input: represents the predictive control input; represents the optimal control input; represents the assumed control input; y des,i (k|t) represents the expected output of the i-th vehicle at time t predicted at time k; h i (v des,i (N p -1|t)) is a function related to fuel economy, v des,i (N p -1|t) indicates that in the prediction time domain N P The expected speed of the i-th vehicle at time t was predicted last time; the local open-loop optimal control objective of the vehicle formation is as follows:

[0054] (1) The control algorithm is recursively feasible for the vehicles in the platoon;

[0055] (2) The platoon is asymptotically stable, that is, when the acceleration of the leading vehicle is zero, all the following vehicles controlled by the distributed controller will reach an equilibrium point over time;

[0056] (3) The convoy has chord stability, that is, the distance error between adjacent vehicles is not amplified by the convoy;

[0057] The objective function of the open-loop optimal control problem is as follows:

[0058]

[0059] The function l i (·) is the cost function associated with vehicle i, defined as

[0060]

[0061] Among them, Q i , R i 、F i , G i , H i is the MPC weight symmetric matrix;

[0062] Step 3: Set constraints on the objective function of the open-loop optimal control problem;

[0063] The constraints are as follows:

[0064]

[0065] in, represents the expected distance vector between vehicle i and vehicle j;

[0066] Step 4: Calculate the optimal speed and acceleration for fuel efficiency:

[0067] The OSQP mathematical solver is used to solve the fuel optimization problem of the continuous quadratic programming algorithm SQP, that is, the optimal fuel consumption driving control algorithm in step 2, to obtain the optimal driving speed and driving acceleration of the pilot vehicle, and the predicted value of the control input is used to update the vehicle state at the k+1th moment. The following vehicle achieves the best fuel consumption according to the driving of the pilot vehicle, thereby achieving the goal of optimal fuel consumption and environmental protection.

[0068] The beneficial effects of adopting the above technical solution are:

[0069] The present invention provides an environmentally friendly driving method for a heterogeneous autonomous driving vehicle formation. Compared with the PF communication topology structure that can only obtain the state information of homogeneous neighboring vehicles in the existing research, the present method adopts the PLF communication topology structure that can simultaneously obtain the state information of the pilot vehicle and the adjacent front vehicle. For the heterogeneous vehicle formation, a distributed model predictive control algorithm (HPDMPC) that can improve the fuel economy of the heterogeneous vehicle formation is designed. At the same time, the algorithm considers the impact of manually driven vehicles on the CAV fleet on the highway, and proves that the method has asymptotic stability. The present method adopts a multi-objective collaborative optimization framework to collaboratively optimize the fuel consumption, acceleration and following distance of each vehicle in the formation, while achieving the goals of reducing fuel consumption, smooth acceleration and deceleration, and meeting the driving goals of safe following distance. A new vehicle formation control HPDMPC algorithm is proposed to achieve fuel efficiency optimization, formation collaborative tracking and stability control. A new distributed constraint condition for the change of following distance is introduced to ensure the recursive feasibility and stability of HPDMPC, as well as the chord stability under the road constraints of the vehicle formation. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 This is a schematic diagram of the longitudinal control of a vehicle formation according to the present invention;

[0071] Figure 2 The vehicle queue of the present invention adopts the PLF communication topology structure;

[0072] Figure 3 It is a trajectory curve diagram of a manually driven vehicle of the present invention;

[0073] Figure 4 It is the control performance curve diagram of the present invention;

[0074] Among them, (a) speed curve, (b) acceleration curve, (c) error diagram between the expected vehicle distance;

[0075] Figure 5 It is a control performance curve diagram of the prior art;

[0076] Among them, (a) speed curve, (b) acceleration curve, (c) error diagram between the expected vehicle distance;

[0077] Figure 6 A curve diagram showing the relationship between the vehicle spacing and fuel consumption of the present invention;

[0078] Figure 7 It is a comparison curve diagram of acceleration, speed and following distance of the pilot vehicle under the control of HPDMPC, CACC and IDM of the present invention;

[0079] Among them, (a) acceleration comparison curve, (b) speed comparison curve, (c) following distance comparison curve;

[0080] Figure 8 This is a flow chart of the HPDMPC control algorithm of the present invention. DETAILED DESCRIPTION

[0081] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0082] A method for environmentally friendly driving of a heterogeneous autonomous driving vehicle formation, comprising the following steps:

[0083] Step 1: Based on the heterogeneous vehicle platooning of the leading vehicle-leader following PLF communication topology, the longitudinal dynamics system of the vehicle platooning, the fuel consumption model and the PLF communication topology are established;

[0084] The heterogeneous vehicle formation consists of a human-driven vehicle (HDV) as an interference vehicle i=0, a pilot vehicle i=1 and follower vehicles i=2,…,7, as shown in Figure 1 As shown;

[0085] The vehicle formation longitudinal dynamics system is specifically:

[0086] This paper improves the classic vehicle nonlinear motion model. This method can more realistically represent heterogeneous nonlinear vehicle dynamics. It also takes into account the influence of time-varying nonlinear uncertainty and air resistance changes during vehicle platooning, and models the longitudinal motion of the vehicle. At time t, the nonlinear model of the vehicle is defined as a one-dimensional coordinate along the direction of the lane to describe the longitudinal motion of the vehicle, and the starting point of the lane is defined as the origin of the coordinates; first, the vehicle driving characteristics under the mutual influence of the front and rear vehicles are described as follows: Let d i represents the following distance, v i (t) and a i (t) represent velocity and acceleration respectively, ε i represents the engine input, represents the speed difference between the i-1th vehicle and the i-th vehicle, It means that the acceleration is obtained by taking the derivative of the velocity of the i-th vehicle. Represents the acceleration a i (t) Derivative to get the jerk The degree of change of reaction acceleration, i is the vehicle number:

[0087]

[0088] f in formula (1) i (v i ,a i ) and g i (v i ,a i ) are:

[0089]

[0090] Different control models proposed in the study of heterogeneous nonlinear vehicle platooning usually ignore the changes in air resistance caused by the "flying geese effect". Therefore, it is necessary to improve the nonlinear model to describe the motion characteristics of each vehicle in the platoon under the influence of air resistance;

[0091]

[0092] where τ i is the dynamic delay coefficient, σ is the air resistance coefficient, m i , are the mass, cross-sectional area, mechanical resistance and drag coefficient of vehicle i, A i is the air resistance matrix associated with vehicle information such as following distance and driving speed; the engine input ε i Expressed as

[0093]

[0094] Among them, ui Represents the control input; converts the nonlinear system into a linear system acceleration derivative Expressed as

[0095]

[0096] Assume that the state of each car is The output is represented as Formula (1) is expressed as

[0097]

[0098] The input matrix Output Matrix State Matrix Where Δt represents the sampling interval;

[0099] Define X(t), Y(t) and U(t) as the state vector, output vector and input vector of each vehicle, namely:

[0100]

[0101] The state vector from the first car to the Nth car is expressed as The output vector from the 1st car to the Nth car is expressed as The input vector from the 1st car to the Nth car is expressed as The dynamic state equation of the vehicle formation is expressed as:

[0102]

[0103] in B′=diag[B1,...,B N ], A′, B′, and C′ represent the vehicle’s state matrix, input matrix, and output matrix, respectively;

[0104] Among them A N represents the state matrix of the Nth vehicle,

[0105] Among them B N represents the input matrix of the Nth vehicle,

[0106] in Represents the N-order unit matrix and output matrix The Kronecker product of

[0107] The control goal of the vehicle formation is to maintain a zero speed difference and a desired safe distance with the vehicle in front, that is,

[0108]

[0109] In summary, the longitudinal dynamic system of the vehicle formation is obtained as shown in equation (6).

[0110] The fuel consumption model is specifically:

[0111] Use the VT-CPFM model to estimate the vehicle's real-time fuel consumption:

[0112]

[0113] Among them, F i (t) represents the instantaneous fuel consumption of vehicle i (L / s), P i (t) is the real-time power of the vehicle (kW), β0, β1, β2 are model parameters; the vehicle power is related to the acceleration and speed, satisfying:

[0114]

[0115] Among them, R i is the vehicle resistance, ρ is the air density, λ i is the rolling friction coefficient, θ(t) is the road slope, η is the vehicle transmission efficiency, is the air resistance coefficient

[0116]

[0117] The PLF communication topology is as follows: Figure 2 As shown, specifically:

[0118] Using directed graphs represents the communication topology of the fleet, is the node set, N is the total number of nodes, It is the communication link between neighboring cars. Contains three attribute matrices: adjacency matrix Laplacian Matrix and traction matrix The adjacency matrix It is used to represent the directional communication between neighboring vehicles and is defined as a ij It is an element in the real number matrix, using 0 or 1 to indicate whether there is a communication link between vehicles, 0 means no link, and 1 means there is a link; Represents a real matrix of order N, namely

[0119]

[0120] in Indicates that there is a directed edge between node j and node i, that is, vehicle i can receive the status information of vehicle j;

[0121] Laplacian Matrix Defined as in, is the in-degree matrix, defined as in, Indicates that node i is The in-degree of

[0122] Traction Matrix Indicates whether there is a communication connection between the following vehicle and the leading vehicle; Wakabe Then p i =1, indicating that there is directed communication between the following vehicle i and the leading vehicle, otherwise, p i = 0. The set indicating whether the pilot vehicle information is transmitted to vehicle i is expressed as

[0123]

[0124] Step 2: Establish a distributed model predictive control algorithm for heterogeneous vehicle platoons; specifically, it includes an optimal fuel consumption driving control algorithm for the lead vehicle and a distributed control algorithm for each following vehicle in the platoon.

[0125] The optimal fuel consumption driving control algorithm for the pilot vehicle is specifically as follows:

[0126] The nonlinear model predictive control is combined with the fuel consumption model. Under physical constraints, the speed information of the preceding vehicle is used to calculate the control input u0 with the best fuel economy. At the same time, in order to avoid collision and consider the effective distance of vehicle-to-vehicle communication, the distance constraint d0 and the vehicle speed constraint v0 are adopted. The vehicle speed constraint v0 is determined by the traffic conditions. To ensure driving smoothness, the vehicle acceleration a0 and the control input u0 need to be constrained:

[0127]

[0128] Where J0 is the prediction interval T P At time k, the predicted state of the pilot car is s, which is the prediction interval range, s∈[0,T P ]; It is about the predicted speed and acceleration at each time step at time k d 0max d 0min are the maximum and minimum following distance limits, d 0，k+s is the following distance of the sth prediction interval at time k; v 0max 、v 0min are the maximum and minimum following speed limits, v 0，k+s is the following vehicle speed in the sth prediction interval at time k; a 0max 、a 0minare the maximum and minimum following acceleration, a 0，k+s is the following vehicle acceleration in the sth prediction interval at time k;

[0129] The distributed control algorithm of each following vehicle in the convoy is:

[0130] Establish the objective function of the open-loop optimal control problem: the following vehicle obtains the information of the preceding vehicle through the V2V communication device and solves the optimal control problem; assuming that there is no communication delay, the same length of prediction range N is used for the output and input vectors of each vehicle p , in the prediction range [t,t+N p ], the following trajectories are defined: represents the output trajectory predicted by the local optimal control problem; It represents the optimal solution to a local problem solved numerically; represents the assumed output trajectory, Indicates that the output trajectory is assumed to be transferred in reverse; define the control input: represents the predictive control input; represents the optimal control input; represents the assumed control input; y des,i (k|t) represents the expected output of the i-th vehicle at time t predicted at time k; h i (v des,i (N p -1|t)) is a function related to fuel economy. The specific expression of this function related to the expected speed is shown in the above formula (8); v des,i (N p -1|t) indicates that in the prediction time domain N P The expected speed of the i-th vehicle at time t was predicted last time; the fuel consumption formula described in formula (15) is expressed in another form The local open-loop optimal control objective of the vehicle platoon is as follows:

[0131] (1) The control algorithm is recursively feasible for the vehicles in the platoon;

[0132] (2) The platoon is asymptotically stable, that is, when the acceleration of the leading vehicle is zero, all the following vehicles controlled by the distributed controller will reach an equilibrium point over time;

[0133] (3) The convoy has chord stability, that is, the distance error between adjacent vehicles is not amplified by the convoy;

[0134] The objective function of the open-loop optimal control problem is as follows:

[0135]

[0136] The function li (·) is the cost function associated with vehicle i, defined as

[0137]

[0138] Among them, Q i , R i 、F i , G i , H i is the MPC weight symmetric matrix: (1) Q i Indicates the error from the expected output, increasing Q i Weights can improve consistency with the desired state. (2) R i Indicates input error, increase R i The weighted vehicle i will travel at a constant speed. (3)F i Indicates the assumed output error, increase F i The weight can make vehicle i maintain the assumed output state. (4)G i represents the error between the assumed outputs of adjacent vehicles, increasing G i The weight can make vehicle i and its neighboring vehicles maintain the same assumed output state. (5)H i Indicates fuel economy, increasing H i The weights can appropriately reduce constraints and reach steady-state fuel consumption as soon as possible, thereby reducing fuel consumption.

[0139] Step 3: Set constraints on the objective function of the open-loop optimal control problem;

[0140] The constraints are as follows:

[0141]

[0142] in, represents the expected distance vector between vehicle i and vehicle j; the terminal constraint (13) aims to ensure that the output of node i at the end of the prediction period is consistent with It is assumed that the average values ​​of the outputs are the same and that node i moves at a constant speed at the end of the prediction horizon without accelerating or decelerating.

[0143] Step 4: Calculate the optimal speed and acceleration for fuel efficiency:

[0144] The OSQP mathematical solver is used to solve the fuel optimization problem of the continuous quadratic programming algorithm SQP, that is, the optimal fuel consumption driving control algorithm in step 2, to obtain the optimal driving speed and driving acceleration of the pilot vehicle, and the predicted value of the control input is used to update the vehicle state at the k+1th moment. The following vehicle achieves the best fuel consumption according to the driving of the pilot vehicle, thereby achieving the goal of optimal fuel consumption and environmental protection.

[0145] In this embodiment, the fuel economy and stability performance of the designed vehicle formation control algorithm are verified. By comparing with the traditional CACC fleet control and the manually driven fleet, the fuel economy and road utilization of the proposed control method are compared. The simulation parameters are shown in Table 4.

[0146] (1) Basic performance and stability verification

[0147] This example mainly uses MATLAB simulation to verify and compare HPDMPC control and CACC to prove their stability and effectiveness. First, the experiment simulates seven heterogeneous connected vehicles using PLF communication topology. The initial state of the pilot vehicle is s0(0) = 0m, v0(0) = 20m / s, and the trajectory of the manually driven vehicle is as follows: Figure 3 As shown in Figure 1, the pilot vehicle is responsible for following and collecting information from the manually driven vehicle, predicting its own acceleration, and transmitting status information to the adjacent following vehicle and each following vehicle in the convoy. The speed, acceleration, and error between the desired vehicle spacing (20 m) of each vehicle in the convoy under HPDMPC control and CACC control are shown in Figure 1. Figure 4 (a) Figure 4 (b) Figure 4 (c) and Figure 5 (a) Figure 5 (b) Figure 5 (c) as shown.

[0148] (2) Analysis of fuel economy and road utilization

[0149] This example uses the EPA US06 driving cycle to simulate the trajectory of the human-driven vehicle (HDV) in front of the convoy. The convoy follows the vehicle in front according to the control targets in equations (7) and (14). To verify the relationship between the following distance and the overall fuel consumption, first, the total fuel consumption under different vehicle spacing is compared, as Figure 6 It can be seen that the smaller the following distance, the smaller the air resistance of the following car, and the lower the overall fuel consumption of the team. Figure 7 As shown in the figure, under the EPA US06 driving condition, the fleet used the traditional CACC control method, manual driving and the HPDMPC algorithm proposed in this paper for comparison. Figure 7 (a) Figure 7 (b) and Figure 7 (c) Represents the acceleration, speed and following distance of the pilot vehicle in HPDMPC, CACC and IDM, respectively. Figure 7 It can be seen that compared with the other two convoy control methods, the convoy acceleration based on the HPDMPC algorithm changes more smoothly and can maintain a more reasonable following distance.

[0150] The fuel efficiency and fuel consumption of each vehicle in the convoy are shown in Table 1 and Table 3 respectively. Compared with CACC, the fuel economy of the leading vehicle using the HPDMPC algorithm is improved by 17.97%, the fuel economy of the convoy as a whole is improved by 19.37%, and the fuel consumption of each following vehicle is also significantly reduced. Figure 8 The pilot vehicle of the HPDMPC design shown in the figure plans its speed through the information of the preceding vehicle, and uses the distance from the preceding vehicle as a buffer to avoid drastic changes in its own speed and the following vehicle's speed. However, in CACC control, the control goal is to maintain a constant following distance and zero speed difference with the preceding vehicle, which will cause the pilot vehicle to make excessive braking and sudden acceleration actions. Such operations will cause frequent changes in engine speed, thereby increasing the overall fuel consumption of the fleet. As can be seen from Table 2, compared with HPDMPC control, although CACC control can make the average following distance smaller and better reduce air resistance, it still cannot significantly reduce fuel consumption under transient traffic.

[0151] Table 1 Fuel economy improvement efficiency

[0152]

[0153] The experimental results show that compared with manual driving, the pilot vehicle under HPDMPC control can ensure a safe distance from the leading vehicle, make the speed curve smoother, and minimize fuel consumption. As shown in Table 1, the fuel economy of the pilot vehicle is improved by 9.08% compared with manual driving. In addition, the fuel economy of the following vehicle using the HPDMPC control method is improved by about 12% to 14%.

[0154] Table 2 Comparison of average vehicle distances in a platoon

[0155]

[0156] There are two main reasons for this: 1) Without previewing highly transient traffic conditions, manually driven vehicle queues cannot effectively plan vehicle movement, so speed fluctuations under transient conditions lead to more fuel consumption. 2) As shown in Table 2, in order to ensure driving safety, drivers keep a long distance from the vehicle in front. However, this driving habit does not effectively utilize the "geese formation effect" in fluid mechanics to reduce the wind resistance of the vehicle behind, which increases driving energy consumption.

[0157] Table 3: Fuel consumption comparison

[0158]

[0159]

[0160] Table 4 Parameters of following vehicles in a vehicle platoon

[0161]

[0162] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) to form a technical solution.

Claims

1. An environmentally friendly driving method for a heterogeneous autonomous driving vehicle formation, characterized in that: The following steps are involved: Step 1: Based on the heterogeneous vehicle platooning of the leading vehicle-leader following PLF communication topology, the longitudinal dynamics system of the vehicle platooning, the fuel consumption model and the PLF communication topology are established; The heterogeneous vehicle formation consists of a manually driven vehicle as an interference vehicle, a pilot vehicle and a plurality of follower vehicles; Step 2: Establish a distributed model predictive control algorithm for heterogeneous vehicle platoons, including an optimal fuel consumption driving control algorithm for the lead vehicle and a distributed control algorithm for each follower vehicle in the platoon. Step 3: Set constraints on the objective function of the open-loop optimal control problem; Step 4: Calculate the optimal speed and acceleration for fuel efficiency: The OSQP mathematical solver is used to solve the fuel optimization problem of the continuous quadratic programming algorithm SQP, that is, the optimal fuel consumption driving control algorithm in step 2, to obtain the optimal driving speed and driving acceleration of the pilot vehicle, and the predicted value of the control input is used to update the vehicle state at the k+1th moment. The following vehicle achieves the best fuel consumption according to the driving of the pilot vehicle, thereby achieving the goal of optimal fuel consumption and environmental protection.

2. The environmentally friendly driving method for heterogeneous autonomous driving vehicle formation according to claim 1, characterized in that: The vehicle formation longitudinal dynamics system described in step 1 is specifically: At time t, the nonlinear model of the vehicle is defined as a one-dimensional coordinate along the lane direction to describe the longitudinal motion of the vehicle, and the starting point of the lane is defined as the origin of the coordinates; first, the driving characteristics of the vehicle under the mutual influence of the front and rear vehicles are described as follows: Let d i represents the following distance, v i (t) and a i (t) represent velocity and acceleration respectively, ε i represents the engine input, represents the speed difference between the i-1th vehicle and the i-th vehicle, It means that the acceleration is obtained by taking the derivative of the velocity of the i-th vehicle. Represents the acceleration a i (t) Derivative to get the jerk The degree of change of reaction acceleration, i is the vehicle number: f in formula (1) i (v i ,a i ) and g i (v i ,a i ) are: Improve the nonlinear model to describe the motion characteristics of each vehicle in platooning affected by air resistance; where τ i is the dynamic delay coefficient, σ is the air resistance coefficient, m i , are the mass, cross-sectional area, mechanical resistance and drag coefficient of vehicle i, A i is the air resistance matrix; the engine input ε i Expressed as Among them, u i Represents the control input; converts the nonlinear system into a linear system acceleration derivative Expressed as Assume that the state of each car is The output is represented as Formula (1) is expressed as The input matrix Output Matrix State Matrix Where Δt represents the sampling interval; Define X(t), Y(t) and U(t) as the state vector, output vector and input vector of each vehicle, namely: The state vector from the first car to the Nth car is expressed as The output vector from the 1st car to the Nth car is expressed as The input vector from the 1st car to the Nth car is expressed as The dynamic state equation of the vehicle formation is expressed as: in B′=diag[B1,...,B N ], A′, B′, and C′ represent the vehicle’s state matrix, input matrix, and output matrix, respectively; A N represents the state matrix of the Nth vehicle, B N represents the input matrix of the Nth vehicle, Represents the N-order unit matrix and output matrix The Kronecker product of The control goal of the vehicle formation is to maintain a zero speed difference and a desired safe distance with the vehicle in front, that is, In summary, the longitudinal dynamic system of the vehicle formation is obtained as shown in equation (6); The fuel consumption model is specifically: Use the VT-CPFM model to estimate the vehicle's real-time fuel consumption: Among them, F i (t) represents the instantaneous fuel consumption of vehicle i, P i (t) is the real-time power of the vehicle, β0, β1, β2 are model parameters; the vehicle power is related to the acceleration and speed, satisfying: Among them, R i is the vehicle resistance, ρ is the air density, λ i is the rolling friction coefficient, θ(t) is the road slope, η is the vehicle transmission efficiency, is the air resistance coefficient The PLF communication topology structure is specifically as follows: Using directed graphs represents the fleet communication topology, is the node set, N is the total number of nodes, It is the communication link between neighboring cars. Contains three attribute matrices: adjacency matrix Laplacian Matrix and traction matrix The adjacency matrix It is used to represent the directional communication between neighboring vehicles and is defined as a ij It is an element in the real number matrix, using 0 or 1 to indicate whether there is a communication link between vehicles, 0 means no link, and 1 means there is a link; Represents a real matrix of order N, namely in Indicates that there is a directed edge between node j and node i, that is, vehicle i can receive the status information of vehicle j; Laplacian Matrix Defined as in, is the in-degree matrix, defined as in, Indicates that node i is in-degree; traction matrix Indicates whether there is a communication connection between the following vehicle and the leading vehicle; Wakabe Then p i =1, indicating that there is directed communication between the following vehicle i and the leading vehicle, otherwise, p i = 0; the set indicating whether the pilot vehicle information is transmitted to vehicle i is expressed as 3. The environmentally friendly driving method for heterogeneous autonomous driving vehicle formation according to claim 1, characterized in that: The optimal fuel consumption driving control algorithm for the pilot vehicle described in step 2 is as follows: The nonlinear model predictive control is combined with the fuel consumption model. Under physical constraints, the speed information of the preceding vehicle is used to calculate the control input u0 with the best fuel economy. At the same time, the distance constraint d0 and the vehicle speed constraint v0 are used to constrain the vehicle acceleration a0 and the control input u0: Where J0 is the prediction interval T P Fuel consumption within the range; at time k, the predicted state of the pilot car is s, which is the prediction interval range, s∈[0,T P ]; It is about the predicted speed and acceleration at each time step at time k d 0max ,d 0min are the maximum and minimum following distance limits, d0,k+s is the following distance of the sth prediction interval at time k; v 0max 、v 0min are the maximum and minimum following speed limits, respectively, v0,k+s is the following speed of the sth prediction interval at time k; a 0max 、a 0min are the maximum and minimum following acceleration, a 0，k+s is the following vehicle acceleration in the sth prediction interval at time k; The distributed control algorithm of each following vehicle in the convoy is: Establish the objective function of the open-loop optimal control problem: the following vehicle obtains the information of the preceding vehicle through the V2V communication device and solves the optimal control problem; assuming that there is no communication delay, the same length of prediction range N is used for the output and input vectors of each vehicle p , in the prediction range [t,t+N p ], the following trajectories are defined: represents the output trajectory predicted by the local optimal control problem; It represents the optimal solution to a local problem solved numerically; represents the assumed output trajectory, Indicates that the output trajectory is assumed to be transferred in reverse; define the control input: represents the predictive control input; represents the optimal control input; represents the assumed control input; y des,i (k|t) represents the expected output of the i-th vehicle at time t predicted at time k; h i (v des,i (N p -1|t)) is a function related to fuel economy, v des,i (N p -1|t) indicates that in the prediction time domain N P The expected speed of the i-th vehicle at time t was predicted last time; the local open-loop optimal control objective of the vehicle formation is as follows: (1) The control algorithm is recursively feasible for the vehicles in the platoon; (2) The platoon is asymptotically stable, that is, when the acceleration of the leading vehicle is zero, all the following vehicles controlled by the distributed controller will reach an equilibrium point over time; (3) The convoy has chord stability, that is, the distance error between adjacent vehicles is not amplified by the convoy; The objective function of the open-loop optimal control problem is as follows: The function l i (·) is the cost function associated with vehicle i, defined as Among them, Q i , R i 、F i , G i , H i is the MPC weight symmetric matrix.

4. The environmentally friendly driving method for heterogeneous autonomous driving vehicle formation according to claim 1, characterized in that: The constraints described in step 3 are as follows: in, represents the expected distance vector between vehicle i and vehicle j.

Images