CN112578672B - Unmanned vehicle trajectory control system and trajectory control method based on chassis nonlinearity - Google Patents

Abstract

The invention discloses a chassis nonlinearity-based unmanned vehicle trajectory control system and a trajectory control method thereof, wherein the trajectory control system comprises: the sensing signal collection module is used for obtaining the running state information of the current vehicle and the environmental vehicle and carrying out signal processing; the driving decision module is used for learning appropriate decision parameter values; the track planning module is used for obtaining a feasible track after optimized planning; the method comprises the following steps of designing a nonlinear vehicle model based on a bicycle model, adapting to the model, using a polynomial model improved by a magic formula, and reducing the number of solution variables by using different discrete time steps of the model and control time steps of control variables in order to reduce the calculation time; the method is suitable for high-level automatic driving vehicles, and aims to effectively improve the self-adaptive capacity of a vehicle system under different driving conditions through an optimization model so as to ensure the safety of the system under the condition of obtaining better driving performance.

Description

Trajectory control system and trajectory control of unmanned vehicle based on chassis nonlinearity method

technical field

The invention relates to the field of unmanned driving technology, in particular to a chassis-based non-linear unmanned vehicle trajectory control system and a trajectory control method thereof.

Background technique

The number of casualties caused by traffic accidents has increased every year for the past few decades. Due to the fast operation and accurate perception of self-driving cars, accidents caused by driver distraction and fatigue can be greatly reduced. In addition to improving safety, autonomous vehicles can greatly improve driving comfort, traffic efficiency, and energy economy. At present, the four main methods of trajectory planning are graph search-based method, incremental search method, curve interpolation method and numerical optimization method. Graph search finds the shortest path by traversing all environment grids. However, the resulting paths are not continuous and thus not suitable for self-driving cars. Incremental search methods allow nodes to randomly extend toward a target in a continuous space. However, the generated trajectories are jerky and usually not optimal. Curve interpolation can generate smooth paths, but the result of trajectory planning depends on global waypoints, and this method is very time-consuming in changing environments. In the prior art, numerical optimization methods are usually used to consider the constraints of the road and the host vehicle. Trajectories are generated from constraints and objective functions. However, the optimization process needs to be performed at each time step, which is time-consuming and requires high-level hardware. Therefore, in the road strength planning process, the efficiency and calculation time must also be considered under the premise of ensuring the accuracy of the model.

Contents of the invention

The present invention proposes a chassis-based non-linear unmanned vehicle trajectory control system and its trajectory control method, which includes a nonlinear vehicle model designed based on a bicycle model, a polynomial model adapted to the model through a magic formula improved, and for To reduce computation time, use different discrete time steps for the model and control time steps for the control variables to reduce the number of solution variables. The trajectory control method of unmanned vehicles based on chassis nonlinearity is suitable for high-level automatic driving vehicles. The goal is to effectively improve the adaptive ability of the vehicle system under different driving conditions through the optimization model, so that the system can obtain better driving performance. It also guarantees safety under certain conditions, and solves the above-mentioned problems existing in existing advanced assisted driving and unmanned driving.

One aspect of the present invention provides an unmanned vehicle trajectory control system based on chassis nonlinearity, including a sensing signal collection module, a driving decision-making module and a trajectory planning module, the sensing signal collection module, the driving decision-making module and the trajectory planning module Together they form a parameterized decision-making framework;

The perception signal collection module is used to obtain the driving state information of the current vehicle and the surrounding vehicle, and perform signal processing;

The driving decision-making module is used to learn appropriate decision-making parameter values;

The trajectory planning module is used to obtain a feasible trajectory after optimized planning.

In the non-linear chassis-based unmanned vehicle trajectory control system of the present invention, further, the sensing signal collection module is used to obtain the lanes, speeds, Acceleration, the vehicle's lane, speed, and the relative distance based on the vehicle's lane, and the driving intention of the surrounding vehicle is obtained through the offset between the surrounding vehicle and its lane centerline or the turn signal information, and the data is collected for subsequent learning of driving decisions train.

In the non-linear chassis-based unmanned vehicle trajectory control system of the present invention, further, the driving decision-making module is used to analyze the relationship between human-like driving behavior and traffic environment, and establish different driving decisions.

In the chassis nonlinear-based unmanned vehicle trajectory control system of the present invention, further, the driving decision-making module is used to establish scene optimization according to the scene, so as to solve the value of decision-making parameters in the current scene.

In the chassis nonlinear-based unmanned vehicle trajectory control system of the present invention, further, the driving decision-making module, for urban working conditions or highway working conditions, the decision parameter values include lane changing and lane keeping behaviors.

In the chassis-based nonlinear driverless vehicle trajectory control system of the present invention, further, the trajectory planning module is used to include the establishment of a nonlinear model and the use of a nonlinear model prediction method in the module.

In the chassis-based non-linear driverless vehicle trajectory control system of the present invention, further, the establishment of the non-linear model is based on vehicle dynamics and kinematics models and based on lateral and longitudinal motion control.

The trajectory control system of the unmanned vehicle based on the chassis nonlinearity of the present invention, further, the nonlinear model prediction method, according to the key decision-making parameters of the decision-making layer, adjusts the optimization target and obtains different terminal constraints; and is not based on the fixed trajectory form In the case of , the trajectory planning is carried out by solving the optimization problem online and performing rolling optimization.

Another aspect of the present invention provides a non-linear driverless vehicle trajectory planning method based on the chassis, which is realized by the non-linear driverless vehicle trajectory control system based on the chassis as described in one aspect of the present invention, including the following steps,

Step 1, the establishment of the controller model:

The establishment of the controller model includes the establishment of the vehicle dynamics model and the establishment of the tire model;

Step 2, the establishment of the optimization problem:

Based on various constraints, the specific solution process is as follows:

First, the input of the driving decision-making module based on the rules of the main vehicle condition, traffic information and route planning information; according to the set rules, it is determined whether to take an obstacle avoidance operation according to the existence of obstacles; after determination, the obstacle position and target The lane will be passed to the trajectory planning module; the trajectory planning module plans the optimal trajectory based on the passed information, and generates a sequence of (x, y, δ, a); converts the sequence into an actuator, and enables automatic driving The car follows the generated optimal trajectory;

Secondly, trajectory planning is formed based on the decision of the upper module, and the trajectory planning algorithm is responsible for planning the trajectory, so that the vehicle performs lane change or lane keeping action; the implementation of trajectory planning is expressed as an optimal control problem in the equation, in order to predict the range [t, t +T] to find the most control sequence of the control variable U=[δ,a] ^T , so as to obtain formulas (6a) and (6b),

Then, as shown in equations (11a) and (11b), the acceleration and steering wheel angle are controlled within the preset range, so that the optimal control result is within the execution range of the vehicle; when an obstacle appears, the vehicle is avoided things. In the optimization process, the obstacle coordinates of each optimization step fall outside the preset range around the host vehicle, so that the planned trajectory avoids collision with obstacles; because the longitudinal velocity of the vehicle is much greater than the lateral velocity, the When driving through obstacles, a larger safety distance is left in the vertical direction than in the horizontal direction; then equation (11c) is used to set the obstacle avoidance.

a _min ≤ a(t) ≤ a _max (11a)

δ _f,min ≤ δ(t) ≤ δ _f,max (11b)

Step 3, quick solution:

Obtain the trajectory directly by optimization, speed up the calculation by simplifying the simulation process, and reduce the number of variables to be solved by extending the control time step.

In the non-linear chassis-based unmanned vehicle trajectory planning method of the present invention, further, the establishment of the controller model in step 1 includes the following steps,

The first step is the establishment of the vehicle model,

The nonlinear model is based on longitudinal and lateral dynamics, with the origin of the vehicle coordinate system fixed at the center of vehicle mass; the x-axis is parallel to the ground and the y-axis points to the left of the driver. The z-axis is perpendicular to the plane formed by the x-axis and the y-axis; according to Newton's second law, formulas (1) and (2) can be obtained,

2·(F _yf +F _yr )＝m·a _y (1)

where m is the mass of the vehicle, I _z is the moment of inertia about the z-axis; F _yf and F _yr are the lateral forces of a single front and rear tire, respectively. a _y represents the lateral acceleration, ω represents the yaw rate of the vehicle body; l _f and l _r represent the distances from the center of mass to the front and rear axles respectively; the side slip angle can be calculated as β=v _y /v _x , where v _y and v _x are respectively are the lateral and longitudinal velocities of the vehicle's center of gravity.

根据坐标系，前轮和后轮侧滑角表示为式(3)和式(4)，Then the sideslip rate can be calculated as

According to the coordinate system, the sideslip angles of the front and rear wheels are expressed as formula (3) and formula (4),

Combining the above dynamics and kinematics relations, the nonlinear control model is expressed as equations (5a) and (5b),

是车辆的状态矢量；

是车辆的航向角；U＝[a,δ]是控制变量的向量；a是纵向加速度，δ是前轮的转向角；in,

is the state vector of the vehicle;

is the heading angle of the vehicle; U=[a,δ] is the vector of control variables; a is the longitudinal acceleration, and δ is the steering angle of the front wheels;

The second step is the establishment of the tire model,

When the lateral acceleration is small, the lateral force of the tire has a linear relationship with the slip angle, expressed as F _y = kα _y , where k is the value of the tire's cornering stiffness; the tire test data is fitted by a combination of trigonometric functions; according to the Different meanings, use the same formula to express longitudinal force, lateral force or alignment torque; the formula is as in formula (7),

In this model, the lateral force is a function of the slip angle, the vertical load of the tire and the camber angle of the tire; fitting equation (7) by a cubic polynomial of the slip angle calculates the same lateral force when the slip angle is opposite absolute value, F _y (α) is an odd function, then when constructing the polynomial fitting formula, the odd power of the slip angle is used, and the formula (8),

F _y =k ₁ α+k ₂ α ³ +k ₃ α ⁵ (8)

Among them, k ₁ , k ₂ , k ₃ are coefficients calculated by numerical fitting method, and α is tire slip angle.

When the body rolls, the vertical load on the tires changes; when the vehicle turns, the flexible components in the steering system deform, and the lateral force of the tires deforms the suspension system, the nonlinearity of the chassis will cause the vehicle model to change, or in When driving at high speed, a coefficient is added before the nonlinear lateral force of the wheel to offset the influence of the chassis nonlinearity on the vehicle model, and formula (9) is obtained.

F _y,non ＝e·(k ₁ α+k ₂ α ³ +k ₃ α ⁵ ) (9)

In the unmanned vehicle trajectory planning method based on chassis nonlinearity of the present invention, further, when the tire model is established in the second step, the lateral force of the front and rear tires is expressed as formula (10a) and (10b) using a polynomial tire model .

In the unmanned vehicle trajectory planning method based on chassis nonlinearity of the present invention, further, when the optimization problem is established in step 2, the absolute value of the vehicle acceleration and the steering angle are reduced, and the number of times of shifting and steering is reduced at the same time; Use the evaluation function to limit the acceleration, steering angle and corresponding derivatives to limit the control quantity and its derivatives, such as formula (12).

In the non-linear chassis-based unmanned vehicle trajectory planning method of the present invention, further, the difference between a and the expected acceleration and the difference between y and the target coordinate are also added to the objective function, wherein the target coordinate is given by Lane change or lane keeping decisions are determined and generated by the decision layer, as in equations (13) and (14),

P _a (X,U,t)=[a(t)-a _tar ] ² (13)

P _y (X,U,t)=[y(t)-y _tar ] ² (14)

Then the cost function L can be expressed as formula (15).

In the non-linear chassis-based unmanned vehicle trajectory planning method of the present invention, further, when solving quickly in step 3, different control variables have different control time steps.

In the non-linear chassis-based unmanned vehicle trajectory planning method of the present invention, further, when solving quickly in step 3, a shorter discrete time step and a smaller control time step are selected, while avoiding the discrete time step of the model. The length and control time step are equal to improve the calculation speed and obtain a reasonable trajectory.

The unmanned vehicle trajectory control system and its trajectory control method based on chassis nonlinearity of the present invention can achieve the following beneficial effects:

The unmanned vehicle trajectory control system and its trajectory control method based on the chassis nonlinearity of the present invention have the following advantages, (1) it uses a nonlinear model that is more accurate than the traditional bicycle model and considers both longitudinal and lateral dynamics . Use the steering and braking system to realize the control of emergency obstacle avoidance; (2) The polynomial model introduced by it can not only be well adapted to the proposed nonlinear vehicle model, but also has the advantages of calculation speed and accuracy; ( 3) It successfully reduces the time of the solution process used by using different discrete time steps of the model and control time steps of the control variables; (4) Its more accurate model, the control sequence can be used as a feed-forward control, and With fine feedback control. This enables the vehicle to achieve good motion control even at high speeds.

Description of drawings

The accompanying drawings described here are used to provide a further understanding of the present invention, and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention, and do not constitute improper limitations to the present invention. In the attached picture:

Figure 1 is a block diagram of the control system.

Figure 2 is a schematic diagram of a simplified vehicle model.

Figure 3 is a comparison of the linear tire model, the magic formula tire model and the polynomial tire model.

Figure 4 shows the optimal trajectory planning hierarchy.

Figure 5 shows the safe distance range of the main vehicle.

detailed description

In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and corresponding drawings. Apparently, the described embodiments are only some of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

The technical solutions provided by various embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

Example 1

The trajectory control system for unmanned vehicles based on chassis nonlinearity includes multiple sub-modules, and its structural block diagram is shown in Figure 1. Framework, suitable for this trajectory control system. Using key parameters instead of human-like driving decisions can better adapt to complex decisions in different driving decisions and various driving scenarios.

Among them, the sensory signal collection module is used to obtain the driving state information of the current vehicle and the surrounding vehicle, and perform signal processing, including: the lane, speed, acceleration, and local The car's lane, speed and relative distance based on the car's lane, and the driving intention of the surrounding car can be obtained through the offset of the surrounding vehicle and its lane centerline or the turn signal information, and the collected data is used for subsequent learning and training of driving decisions.

Among them, the driving decision module learns appropriate decision parameter values. The key parameter set of driving decision is established by analyzing the relationship between human-like driving behavior and traffic environment, and the characteristics of different driving decisions. The optimization problem of different decision-making scenarios is to establish the corresponding optimization problem according to the characteristics of the scenario, and finally solve the key decision-making parameter values in the current scenario. For urban conditions, or highway conditions, the most common behaviors are lane changing and lane keeping. The characteristic relationship of its decision parameters is simple and consistent.

The trajectory planning module is used to obtain the feasible trajectory after optimized planning. Included in the module is the establishment of nonlinear models and the use of nonlinear model forecasting methods. Among them, the nonlinear model is established based on the vehicle dynamics and kinematics model and considers the lateral and longitudinal motion control. The use of nonlinear model prediction method, according to the key decision-making parameters of the decision-making layer, adjusts the optimization goal and obtains different terminal constraints; and without being based on a fixed trajectory form, the trajectory planning is carried out by solving the optimization problem online and rolling optimization.

Example 2

The unmanned vehicle trajectory planning method based on chassis nonlinearity, it is realized by the unmanned vehicle trajectory control system based on chassis nonlinearity as described in embodiment 1, comprises the following steps,

Step 1, the establishment of the controller model:

The establishment of the controller model in this paper includes the establishment of the vehicle dynamics model and the establishment of the tire model. In the related research of trajectory planning, bicycle model is widely used as a simplified vehicle dynamics model. When using this model, the following assumptions should be followed:

1. Neglecting the influence of the steering system, the steering angle is used as an input to the system.

2. The carriage moves parallel to the ground.

3. The forward speed of the vehicle remains constant.

4. Ignore the influence of side wind.

5. The coefficient of adhesion between the tire and the road is 1.

However, during the planning process, the speed variation of the car also needs to be considered. Therefore, existing bicycle models need to be modified. In the controller model proposed in this paper, the bicycle model is changed to a nonlinear model. The nonlinear model takes into account both longitudinal and transverse dynamics. Compared with the traditional bicycle model, the motion state of the vehicle can be described more accurately.

For any vehicle dynamics model, the tire dynamics model is the most important foundation, which directly affects the accuracy of vehicle dynamics modeling. In most studies, the lateral force on the tire is usually expressed in a linear form. However, this linear relationship no longer exists when the lateral acceleration becomes large. If the above-mentioned nonlinear vehicle model continues to use the lateral force output by the linear tire model, it will cause large errors, so it is necessary to use a more accurate tire model.

1. Establishment of vehicle model

The nonlinear model takes into account both longitudinal and transverse dynamics. The vehicle coordinate system is shown in Figure 2. The entire coordinate system is right-handed, while its origin is fixed at the center of the vehicle mass. The x-axis is parallel to the ground and the y-axis points to the left of the driver. The z axis is perpendicular to the plane formed by the x and y axes. According to Newton's second law, formulas (1) and (2) are obtained

2·(F _yf +F _yr )＝m·a _y (1)

where m is the mass of the vehicle and Iz is the moment of inertia about the _z -axis. F _yf and F _yr are the individual front and rear tire lateral forces, respectively. a _y represents the lateral acceleration, and ω represents the yaw rate of the vehicle body. l _f and l _r represent the distance from the center of mass to the front and rear axes, respectively. The slip angle can be calculated as β=v _y /v _x , where v _y and v _x are the lateral and longitudinal velocities of the vehicle's center of gravity, respectively.

根据坐标系，前轮和后轮侧滑角可以表示为式(3)和(4)，Therefore, the sideslip rate can be calculated as

According to the coordinate system, the sideslip angles of the front and rear wheels can be expressed as equations (3) and (4),

Combining the above dynamics and kinematics relations, the nonlinear control model can be expressed as equations (5a) and (5b),

是车辆的状态矢量。

是车辆的航向角。U＝[a,δ]是控制变量的向量。a是纵向加速度，δ是前轮的转向角。here,

is the state vector of the vehicle.

is the heading angle of the vehicle. U=[a,δ] is a vector of control variables. a is the longitudinal acceleration and δ is the steering angle of the front wheels.

The above model is relatively simple and easy to calculate, but the model does not consider the nonlinearity of the chassis. When the vehicle speed increases, the vehicle gradually enters the nonlinear region, and the error of the traditional vehicle model is very serious. Chassis nonlinearity mainly includes tire nonlinearity, vehicle roll, steering system deformation and suspension system deformation. As the body rolls, the vertical load on the tires will change. When a vehicle corners, the flexible components in the steering system deform. In addition, the lateral forces of the tires can deform the suspension system. The aforementioned chassis nonlinearities can cause significant changes to the vehicle model, especially at high speeds. It is important to incorporate chassis nonlinearities into the vehicle model to improve accuracy without overcomplicating the calculations. Compared with the bicycle model, the form of the vehicle model considering the nonlinearity of the chassis has changed greatly.

2. Establishment of tire model

When the lateral acceleration is small, the lateral force of the tire has a linear relationship with the slip angle, which can be expressed as F _y = kα _y , where k is the value of the tire's cornering stiffness. In most studies, lateral forces are expressed in linear form. However, this linear relationship no longer exists when the lateral acceleration becomes large. It is necessary to use a more accurate tire model. The Magic Formula, introduced by HB Pacejka, is a tire model known for its high precision and widely used in the automotive industry. This formula uses a combination of trigonometric functions to fit tire test data. Depending on what x means, the same formula can be used for longitudinal force, lateral force, or alignment torque. This formula is usually expressed as equation (7),

In this model, the lateral force is a function of the slip angle, the vertical load of the tire and the camber angle of the tire. This curve shows that the lateral force does not have a linear relationship with the slip angle as the slip angle increases. Continuing to use a linear model in this case will produce errors that cannot be ignored. When the car is cornering sharply, there is a lot of lateral acceleration, and the wheels can easily go beyond the linear zone. Therefore, using a linear model in this case produces a large error. To account for the nonlinearity, the tire lateral force in the vehicle model adopted in this paper is fitted with the magic formula, and simulations are carried out below using all three models mentioned above to verify the superiority of the nonlinear model by using the nonlinear model. However, the form of the magic formula is too complex to be used directly. By fitting the magic formula with a cubic polynomial of the slip angle, accuracy is ensured and calculations are greatly reduced. In order to calculate the same absolute value of lateral force at opposite slip angles, F _y (α) must be an odd function. Therefore, when constructing the polynomial fitting formula, the odd power of the slip angle is used to obtain formula (8).

F _y =k ₁ α+k ₂ α ³ +k ₃ α ⁵ (8)

Here, k ₁ , k ₂ , and k ₃ are coefficients calculated by a numerical fitting method, and α is the tire slip angle.

As the body rolls, the vertical load on the tires will change. When a vehicle corners, the flexible components in the steering system deform. In addition, the lateral forces of the tires can deform the suspension system. The aforementioned chassis nonlinearities can cause significant changes to the vehicle model, especially at high speeds. It is important to introduce chassis nonlinearities into the vehicle model to improve accuracy without overcomplicating the calculations. According to the previous work, adding a coefficient in front of the nonlinear lateral force of the wheel can offset the influence of the chassis nonlinearity on the vehicle model, as shown in equation (9).

F _y,non ＝e·(k ₁ α+k ₂ α ³ +k ₃ α ⁵ ) (9)

A nonlinear optimal controller with a nonlinear model was introduced above. The lateral force of the front and rear tires can be expressed as equations (10a) and (10b) using the polynomial tire model,

度时，线性轮胎模型是完全不合适的，而多项式轮胎模型仍然非常接近魔术公式。在装有

Core^TMi7-6700HQ 2.60GHz CPU的笔记本电脑上进行仿真，并使用MATLABR2018a计算1000组横向力。使用多项式轮胎模型的计算比使用魔术公式轮胎模型的计算快74.18％。Figure 3 shows the comparison of linear tire model, magic formula tire model and polynomial tire model. When the side slip angle is greater than

, the linear tire model is completely inappropriate, while the polynomial tire model is still very close to the magic formula. equipped with

The simulation is performed on a laptop with a Core ^TM i7-6700HQ 2.60GHz CPU, and 1000 sets of lateral forces are calculated using MATLABR2018a. The calculation using the polynomial tire model is 74.18% faster than the calculation using the magic formula tire model.

Step 2, the establishment of the optimization problem:

After proposing a suitable controller model, the focus of this paper is to optimize the path planning problem of unmanned vehicles through this model, hoping to obtain a safe, smooth and comfortable trajectory. Therefore, it needs to be solved based on various constraints. The specific solution process is as follows:

The optimal trajectory planning hierarchy is shown in Fig. 4. The host vehicle status, traffic information and route planning information are the inputs of the rule-based decision module. According to the set rules, it will determine whether to take obstacle avoidance operation according to the existence of obstacles. Once determined, obstacle locations and target lanes will be passed to the trajectory planning module. The trajectory planning module is able to plan the optimal trajectory based on the passed information. A sequence of (x, y, δ, a) will be generated. This sequence is translated into actuators, and the self-driving car can follow the generated optimal trajectory.

The formulation of the trajectory planning problem is based on the decisions of the upper modules, and the trajectory planning algorithm is responsible for planning the trajectory so that the vehicle can perform lane changing or lane keeping maneuvers smoothly. The implementation of trajectory planning can express the problem as the optimal control problem in the equation to find the optimal control sequence of the control variable U=[δ,a] ^T in the prediction range [t, t+T], so that the formula ( 6a) and (6b),

This is an optimization problem with both constraints and an objective function. To ensure safety and smoothness, some hard constraints are necessary. As shown in equations (11a) and (11b), the acceleration and steering wheel angle need to be controlled within a certain range to ensure that the best control results are within the vehicle's executable range. Exceeding this range may be detrimental to driving safety and affect comfort, and it is also necessary to avoid such as wheel slippage. In addition, the vehicle must also be able to avoid obstacles when they appear. During optimization, it must be ensured that the coordinates of obstacles at each optimization step fall outside a certain range around the host vehicle to ensure that the planned trajectory does not collide with obstacles. Since the longitudinal speed of the vehicle is much greater than the lateral speed, when avoiding obstacles, a larger safety distance must be left longitudinally than laterally. Therefore, equation (11c) is used to set the obstacle avoidance, Fig. 4 shows the safe distance when performing lane change and lane keeping, and Fig. 5 shows the safe distance range of the ego vehicle.

a _min ≤ a(t) ≤ a _max (11a)

δ _f,min ≤ δ(t) ≤ δ _f,max (11b)

In addition, the planned trajectory also needs to ensure high efficiency and good comfort, so these factors need to be added to the evaluation function. In order to ensure comfort as much as possible, it is necessary to reduce the absolute value of vehicle acceleration and steering angle, while reducing the number of shifts and steering can greatly improve comfort. Therefore, it is necessary to limit the control quantity and its derivatives by using the evaluation function to limit the acceleration, steering angle and corresponding derivatives, as shown in Equation (12).

For efficiency, the difference between a and the desired acceleration and the difference between y and the target coordinate, where the target coordinate is determined by the lane change or lane keeping decision, is also added to the objective function. Generated by the decision-making layer, such as formula (13), (14) and (15).

P _a (X,U,t)=[a(t)-a _tar ] ² (13)

P _y (X,U,t)=[y(t)-y _tar ] ² (14)

Therefore, the cost function L can be expressed as:

Step 3, quick solution:

In this work, trajectories are obtained directly through optimization, so the computation time should be considered. In order to speed up the calculation, some simplifications are made in the simulation. First, the model requires small discrete time steps to ensure convergence, and small control variable steps will greatly increase computation time when the control time steps are as long as the discrete steps. On the control side, the control variables don't actually need to change as fast as the model discretization. Extending the control time step can greatly reduce the number of variables to be resolved. In addition, different control variables can also have different control time steps, taking into account the different characteristics of the control variables.

Therefore, by comparing the effect of the discrete time step t _d of the model and the control time steps t _c1 and t _c2 of the control variables u1 and u2 , in general, shorter time steps have more stable computation times, but The discrete time step size of the model is equal to the control time step size, which can significantly increase the calculation time. Choosing a shorter discrete time step and a smaller control time step while avoiding the model discrete time step being equal to the control time step can greatly increase the calculation speed and still obtain a reasonable trajectory.

The above descriptions are only examples of the present invention, and are not intended to limit the present invention. Various modifications and variations of the present invention will occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the scope of the claims of the present invention.

Claims (6)

1. A locus planning method of an unmanned vehicle based on chassis nonlinearity adopts a locus control system, and is characterized in that the locus control system comprises a sensing signal collection module, a driving decision module and a locus planning module, and the sensing signal collection module, the driving decision module and the locus planning module jointly form a parameterization decision-making frame;

the sensing signal collection module is used for obtaining the running state information of the current vehicle and the environmental vehicle and carrying out signal processing;

the driving decision module is used for learning a proper decision parameter value;

the track planning module is used for obtaining a feasible track after optimized planning;

the perception signal collection module is used for obtaining the driving intention of the environmental vehicle through the deviation of the environmental vehicle and the lane central line of the environmental vehicle or the steering lamp information by means of the lane, the speed and the acceleration of the surrounding vehicle, the lane and the speed of the vehicle and the relative distance taking the lane of the vehicle as the reference, which are obtained by a radar environment perception element and a vehicle-mounted camera in the vehicle-mounted intelligent perception module, and collecting data for the learning and training of the subsequent driving decision;

the driving decision module is used for analyzing the relation between the human-like driving behaviors and the traffic environment and establishing different driving decisions;

the trajectory planning method comprises the following steps,

step one, establishing a controller model:

the establishment of the controller model comprises the establishment of a vehicle dynamic model and the establishment of a tire model;

step two, establishing an optimization problem:

solving based on each constraint condition, wherein the concrete solving process is as follows:

first, a host vehicle condition, traffic information, and route planning information are input to the driving decision module based on rules; determining whether to adopt obstacle avoidance operation according to the existence of the obstacle according to a set rule; after the determination, transmitting the position of the obstacle and the target lane to the trajectory planning module; the track planning module plans an optimal track based on the transmitted information and generates a sequence of (x (t), y (t), delta (t) and a (t)), wherein x (t) is a vehicle abscissa at the moment t, y (t) is a vehicle ordinate at the moment t, delta (t) is a front wheel rotation angle at the moment t, and a (t) is a vehicle longitudinal acceleration at the moment t; converting the sequence into an actuator and enabling the automatic driving automobile to follow the generated optimal track;

secondly, a track planning is formed based on the decision of the superior module, and a track planning algorithm is responsible for planning the track, so that the vehicle performs lane change or lane keeping action; the trajectory planning implementation is expressed as an optimal control problem in equations to predict the range [ T, T + T [ ]]Find the control variable U (t) = [ delta (t), a (t)] ^T Thereby obtaining equations (6 a) and (6 b),

minJ＝∫ _t ^t+T L(X(t),U(t),t)dt (6a)

where m is the mass of the vehicle, I _z Is the moment of inertia about the z-axis; f _yf And F _yr Lateral forces of a single front tire and rear tire, respectively; ω represents the yaw rate of the vehicle body; l. the _f And l _r Respectively representing the distance from the centroid to the front-rear axis; the centroid slip angle can be calculated as β = v _y /v _x Wherein v is _y And v _x Transverse and longitudinal respectively of the centre of gravity of the vehicleA speed;

is the heading angle of the vehicle at time t;

then, as shown in equations (11 a) and (11 b), the acceleration and the steering wheel angle are controlled within the preset ranges so that the optimum control result is within the execution range of the vehicle; when an obstacle appears, the vehicle avoids the obstacle; during the optimization, the coordinates of the obstacles of each optimization step are made to fall outside a preset range around the host vehicle, so that the planned trajectory avoids colliding with the obstacles; because the longitudinal speed of the vehicle is far greater than the transverse speed, when the vehicle avoids an obstacle, a larger safety distance is reserved in the longitudinal direction than in the transverse direction; equation (11 c) is used to set the avoidance of the obstacle;

a _min ≤a(t)≤a _max (11a)

δ _f,min ≤δ(t)≤δ _f,max (11b)

wherein, a _min ,a _max Respectively, the minimum value and the maximum value of the longitudinal acceleration of the vehicle, a (t) is the longitudinal acceleration of the vehicle at the moment t, delta _f,min ，δ _f,max The minimum and maximum values of the control quantity front wheel rotation angle are respectively, delta (t) is the control quantity front wheel rotation angle at the moment t, x (t), x _ob (t),x _safe Respectively, the abscissa position of the vehicle at the time t, the abscissa position of the obstacle vehicle, and the safety distance, y (t), left by the abscissas in the case of no collision _ob (t)，y _safe Respectively representing the vertical coordinate position of the vehicle at the moment t, the vertical coordinate position of the obstacle vehicle and the safety distance reserved by the vertical coordinate under the condition of no collision;

step three, fast solving:

the trajectory is directly obtained by optimization, the calculation speed is increased by simplifying the simulation process, and the number of variables to be solved is reduced by prolonging the control time step length.

2. The chassis nonlinearity-based unmanned vehicle trajectory planning method of claim 1, wherein the controller model of step one is created comprising the steps of,

the first step, the building of the vehicle model,

the nonlinear model is based on longitudinal and transverse dynamics, and the origin of a vehicle coordinate system is fixed at the center of the vehicle mass; the x-axis is parallel to the ground, and the y-axis points to the left of the driver; the z-axis is perpendicular to the plane formed by the x-axis and the y-axis; the equations (1) and (2) can be obtained according to Newton's second law,

2·(F _yf +F _yr )＝m·a _y (1)

where m is the mass of the vehicle, I _z Is the moment of inertia about the z-axis; f _yf And F _yr Lateral forces of a single front tire and rear tire, respectively; a is _y Represents the lateral acceleration, ω represents the yaw rate of the vehicle body; l _f And l _r Respectively representing the distance from the centroid to the anterior-posterior axis; the centroid slip angle can be calculated as β = v _y /v _x Wherein v is _y And v _x Lateral and longitudinal velocities of the vehicle's center of gravity, respectively;

the side slip rate can be calculated as

The front and rear wheel slip angles are expressed as equations (3) and (4) according to a coordinate system,

in combination with the above kinetic and kinematic relationships, the nonlinear control model is represented by equations (5 a) and (5 b),

wherein,

is the heading angle of the vehicle at time t;

the second step, the building of the tire model,

when the lateral acceleration is small, the lateral force of the tire is linear with the lateral slip angle, denoted as F _y ＝kα _y Wherein k is the cornering stiffness value of the tyre;

wherein X denotes a tangent of a tire side slip, Y (α) denotes a tire side force, α denotes a tire side slip angle, and S _h Is a horizontal shift of the curve, S _v The vertical direction drift of the curve, C is the shape factor of the curve, D is the peak factor of the curve, B is the rigidity factor, and E is the curvature factor of the curve;

in this model, the tire side force is a function of the tire slip angle, the vertical load of the tire and the tire camber angle, equation (7) is fitted by a cubic polynomial of the tire slip angle, and the same absolute value of the side force, F, is calculated when the tire slip angles are opposite _y For an odd function, when a polynomial fitting formula is constructed, the odd power of the tire slip angle is adopted to obtain a formula (8),

F _y ＝k ₁ α+k ₂ α ³ +k ₃ α ⁵ (8)

wherein k is ₁ ,k ₂ ,k ₃ Is a coefficient calculated by a numerical fitting method, and alpha is a tire slip angle;

as the vehicle body rolls, the vertical load on the tires changes; when the vehicle turns, a flexible part in a steering system deforms, and a suspension system deforms due to the lateral force of tires, so that the non-linearity of a chassis can cause the change of a vehicle model, or when the vehicle runs at a high speed, a coefficient e is added in front of the non-linear lateral force of wheels to counteract the influence of the non-linearity of the chassis on the vehicle model, and an equation (9) is obtained;

F _y,non ＝e·(k ₁ α+k ₂ α ³ +k ₃ α ⁵ ) (9)

wherein, F _y,non Is the nonlinear lateral force of the wheel.

3. The method of chassis nonlinearity based unmanned vehicle trajectory planning according to claim 2,

in the second step of building the tire model, the lateral forces of the front and rear tires are expressed as equations (10 a) and (10 b) using a polynomial tire model;

wherein k is _f1 ，k _f2 ，k _f3 Is the lateral force fit coefficient for the front wheel obtained as described in equation (8);

k _r1 ，k _r2 ，k _r3 is the lateral force fit coefficient for the rear wheel obtained as described in equation (8).

4. The method of chassis nonlinearity based unmanned vehicle trajectory planning according to claim 1,

when the optimization problem is established in the second step, the absolute value of the acceleration of the vehicle and the steering angle are reduced, and meanwhile, the times of gear shifting and steering are reduced; the control quantity and its derivatives are limited by limiting the acceleration, steering angle and corresponding derivatives using an evaluation function, as in equation (12):

wherein, P _loss (X (t), U (t), t) are evaluation functions,

the derivative of the longitudinal acceleration of the vehicle corresponding to the angle of rotation of the front wheels.

5. The chassis nonlinearity based unmanned-vehicle trajectory planning method of claim 4,

a (t) and a desired acceleration a _tar The difference between and y (t) and the target coordinate y _tar The difference between them is also added to the objective function, where the target coordinates are determined by the lane change or lane keeping decision and generated by the decision layer, as equations (13) and (14),

P _a (X(t),U(t),t)＝[a(t)-a _tar ] ² (13)

P _y (X(t),U(t),t)＝[y(t)-y _tar ] ² (14)

the cost function L can be expressed as equation (15),

6. the chassis-nonlinearity-based unmanned-vehicle trajectory planning method of claim 1, wherein different control variables have different control-time steps when solved for fast in step three.

Images