Vehicle Dyanmics and Control 2. Dynamic bicycle model

参考链接如下：
Vehicle Dyanmics and Control

2. Dynamic bicycle model

2.1 Review: Dynamics of a rigid body

Translatory motion

Consider a point P with mass m in ${\mathbb{R}}^3$. Let $\overrightarrow{r_P}(t)\in {\mathbb{R}}^3$ be its position in an inertial reference frame, and $\overrightarrow{v_P}(t)\in {\mathbb{R}}^3$ its velocity, and $\overrightarrow{a_P}(t)\in {\mathbb{R}}^3$ its acceleration.

• The (linear) momentum of P is defined as $\overrightarrow{p_P}(t)=m(t)\overrightarrow{v_P}(t)$.
• By Newton’s second law, $\frac{d\overrightarrow{p_P}(t)}{dt}={\sum }_i\overrightarrow{F_i}(t)$, or if m(t) is constant. $m\overrightarrow{a_P}(t)={\sum }_i\overrightarrow{F_i}(t)$, where $\overrightarrow{F_i}(t)$ represent all forces acting on the point mass P.
  
  Consider a rigid body B with mass m in ${\mathbb{R}}^3$. Let $\overrightarrow{r_c}(t)\in {\mathbb{R}}^3$ be the position of its center of gravity C in an inertial reference frame, and $\overrightarrow{v_C}(t)\in {\mathbb{R}}^3$ its velocity, and $\overrightarrow{a_C}(t)\in {\mathbb{R}}^3$ its acceleration.
  
• The (linear) momentum of B is defined as $\overrightarrow{p_B}(t)=m(t)\overrightarrow{v_C}(t)$.
• By the center-of-gravity principle, the center of gravity of a rigid body behaves like a point mass with mass m and as if all forces acting anywhere on the rigid body were acting directly on this point mass: $\frac{d\overrightarrow{p_B}(t)}{dt}={\sum }_i\overrightarrow{F_i}(t)$, or if m(t) is constant. $m\overrightarrow{a_C}(t)={\sum }_i\overrightarrow{F_i}(t)$, where $\overrightarrow{F_i}(t)$ represent all forces acting on the rigid body B.
  

Rotatory motion

Consider a rigid body B with mass m in ${\mathbb{R}}^3$. For the rotatory motion, also the geometric shape of B and the spatial distribution of its mass is important.

• Let the reference point C be the body’s center of gravity. (Alternatively, C may be chosen as a fixed point in the body frame that is also a fixed point in the inertial frame.)
• Let $\rho (x,y,z)$ be the body’s density function, so that $m={\int }_{B}dm={\int }_B\rho (x,y,z)dxdydz$.
  
  
• The inertia tensor of B is defined as
  
• Let $\overrightarrow{\omega }$ be the vector of angular velocities of the rigid body about the x-, y-, z-axes:
  $$\overrightarrow{\omega }=\left[\begin{array}{c}{\omega }_x\\ {\omega }_y\\ {\omega }_z\end{array}\right]$$
• The angular momentum ${\overrightarrow{L}}^{(c)}$ of the rigid body (w.r.t. C) is given by ${\overrightarrow{L}}^{(c)}=\Theta \overrightarrow{\omega }$.
  
  
• By the angular momentum principle, $\frac{d}{dt}{\overrightarrow{L}}^{(c)}(t)={\sum }_i\overrightarrow{M_i^{(c)}}(t)$, or if $\Theta (t)$ is constant. $\Theta \dot{\overrightarrow{\omega }}(t)={\sum }_i\overrightarrow{M_i^{(c)}}(t)$, where $\overrightarrow{M_i^{(c)}}(t)$ are the momentums of all forces acting on B w.r.t. C.
  
  

2.2 Tires Terminology and basics

2.3 Lateral tire models

The purpose of tire models is to determine the forces and moments that act on the vehicle via its tire-road contact.

In the contact patches, tire forces and tire moments are acting on each tire:

Tire forces	Tire moments
vertical force ${\overrightarrow{F}}_z$	aligning torque ${\overrightarrow{M}}_z$ （回正力矩）（主要考虑）
longitudinal force ${\overrightarrow{F}}_x$	overturning moment ${\overrightarrow{M}}_x$ （倾覆力矩）
lateral force ${\overrightarrow{F}}_y$	rolling resistance moment ${\overrightarrow{M}}_y$ （滚动阻力矩）
Assumption: For lateral tire models, the longitudinal force ${\overrightarrow{F}}_x$ is neglected (i.e. no acceleration and braking).
Difficulties:

• Uneven pressure distribution in the tire contact patch
• Complexity of the tire-road contact
• Different road and tire conditions

Analytic tire models

The main idea is to derive the tire forces from first principles in mechanics, in particular Coulomb friction (adhesion + friction) in the contact patch and Elastic deformation of the tire.

The following elastic Foundation model for lateral tire forces has been developed by Ernst Fiala in 1954:

In the simplest case, we may allow for a discontinuous displacement of the center line and a discontinuous slope:

Case 1: The maximum static friction force of $\mu F_z$ ($\mu$: static friction parameter) is not exceeded anywhere in the contact patch.
Lateral tire force: $F_y=\frac{1}{2a}{\int }_{x=0}^{x=2a}c(x)y(x)dx=\frac{1}{2a}{\int }_{x=0}^{x=2a}c\cdot x\tan \alpha dx=ca\tan \alpha \approx ca\alpha$. Cornering stiffness $c_{\alpha }=ca$, so $F_y=c_{\alpha }\alpha$, this tire model is also called the linear tire model.
Aligning torque: $M_z=\frac{1}{2}{\int }_{x=0}^{x=2a}c(x)y(x)(x-a)dx=\frac{1}{2a}{\int }_{x=0}^{x=2a}c\cdot x\tan \alpha (x-a)dx\approx (\frac{1}{3}c{a}^2)\alpha$

Case 2: The slip angle $\alpha$ is large enough that the maximum static friction $\mu F_z$ is exceeded at some place in the contact patch.
The center line displacement is given by
$$\begin{gathered} y(x)=x\tan \alpha if0<x<\bar{x}(staticfriction) \\ y(x)=\frac{\mu F_z}{c}if\bar{x}\le x\le 2a(dynamicfriction) \\ \bar{x}\tan \alpha =\frac{\mu F_z}{c}\Rightarrow \bar{x}=\frac{\mu F_z}{c\tan \alpha } \end{gathered}$$
Lateral tire force: $F_y=\frac{1}{2a}[{\int }_{x=0}^{x=\bar{x}}cx\tan \alpha dx+{\int }_{x=\bar{x}}^{x=2a}c\frac{\mu F_z}{c}dx]=\mu F_z-\frac{{\mu }^2{F}_z^2}{4ac\tan \alpha }$, if $\alpha \ge \frac{\mu F_z}{2ac}$.
Aligning torque: $M_z=\frac{1}{2a}[{\int }_{x=0}^{x=\bar{x}}cx\tan \alpha (x-a)dx+{\int }_{x=\bar{x}}^{x=2a}c\frac{\mu F_z}{c}(x-a)dx]=\frac{{\mu }^2{F}_z^2}{4c\tan \alpha }-\frac{{\mu }^3{F}_z^3}{16c^{2}a{\tan }^2\alpha }$, if $\alpha \ge \frac{\mu F_z}{2ac}$.

Empirical tire models

The main idea is to fit the parameters of predefined functions to measured data from a tire test bench. Empirical data shows that the lateral tire force $F_y$ roughly behaves as follows:

The shape of the tire force curve depends on many variables, e.g.:

• size of the tire;
• construction type (radial-ply, …);
• number and angle of plies;
• general tire / tread design;
• aspect ratio;
  For any given tire (above variables fixed), the main factors are:
• tire inflation pressure;
• tire load (vertical force $F_z$;
  A frequent assumption is that $\frac{F_y}{F_z}$ follows the same curve, but actually:
  
  The Pacejka tire model is the most commonly used empirical tire model practice. It is based on the Magic Formula:
  $$F_y=D\sin [C\arctan (B\alpha -E(B\alpha -E(B\alpha -\arctan (B\alpha ))))]\mu F_z$$

Variables	Meaning
$\alpha$	tire side slip angle
$\mu$	tire-road friction coefficient
$F_z$	vertical tire force
B	stiffness factor
C	shape factor
D	peak value
E	curvature factor

2.4 The dynamic bicycle model with linear tires

Assumptions:

• wheels of the front and rear axle are lumped together into one front and one rear wheel;
• reference point C is the vehicle’s center of gravity;
• vehicle’s motion is restricted to the X, Y-plane;
• for the lateral and yaw motion, the vehicle is considered as a rigid body, whose dynamics are determined by the fundamental laws of motion;
• only lateral tire forces, generated by a linear tire model;
• the steering angle is small: $\sin \delta \approx \delta$, $\tan \delta \approx \delta$, $\cos \delta \approx 1$;
• for the longitudinal motion, the velocity $v_{lon}$ is constant;

Laws of motion

Lateral dynamics:
$$\begin{gathered} m{a}_y^{(x,y)}={\dot{v}}_{lat}+\omega v_{lon}=\sum _i{F}_{y,i}=F_r+F_f\cos \delta \approx F_r+F_f \\ {a}_y^{(x,y)}={\dot{v}}_{lat}+\omega v_{lon}(\mathrm{centripetal}\mathrm{acceleration}) \\ \Rightarrow m({\dot{v}}_{lat}+\omega v_{lon})=F_r+F_f\text{\hspace{1em}(1)} \end{gathered}$$
Yaw dynamics:
$$\begin{gathered} I_z\dot{\omega }=\sum _i{M}_I^{(c)}=-l_r{F}_r+l_f{F}_f\cos \delta \\ \Rightarrow I_z\dot{\omega }=-l_r{F}_r+l_f{F}_f\text{\hspace{1em}(2)} \end{gathered}$$

Tire forces

Linear tire model: $F_f=-c_{\alpha ,f}{\alpha }_f$, $F_r=-c_{\alpha ,r}{\alpha }_r$
Velocity geometry: ${\alpha }_f\approx \tan {\alpha }_f=\frac{v_{A,\eta }}{A,ϵ}$, ${\alpha }_r\approx \tan {\alpha }_r=\frac{v_{B,y}}{B,x}$
Kinematics: $v_{B,x}=v_{lon}$, $v_{B,y}=v_{lat}-\omega l_r$, $v_{A,x}=v_{lon}$, $v_{A,y}=v_{lat}+\omega l_f$
Coordinate transformation $(x,y)\to (ϵ,\eta )$:
$$\begin{gathered} v_{A,ϵ}=v_{A,x}\cos \delta +v_{A,y}\sin \delta \\ {v}_{A,\eta }=-v_{A,x}\sin \delta +v_{A,y}\cos \delta \end{gathered}$$
$$\begin{gathered} \Rightarrow F_f=-c_{\alpha ,f}{\alpha }_f=-c_{\alpha ,f}\frac{v_{A,\eta }}{A,ϵ}=-c_{\alpha ,f}\frac{-v_{lon}\delta +v_{lat}+\omega l_f}{v_{lon}+(v_{lat}+\omega l_f)\delta }\approx c_{\alpha ,f}\delta -c_{\alpha ,f}\frac{v_{lat}+\omega l_f}{v_{lon}} \\ \mathrm{if}{v}_{lon}\gg (v_{lat}+\omega l_f)\delta \text{\hspace{1em}(3)} \end{gathered}$$
$$F_r=-c_{\alpha ,r}{\alpha }_r=-c_{\alpha ,r}\frac{v_{B,y}}{B,x}=-c_{\alpha ,r}\frac{v_{lat}-\omega l_r}{v_{lon}}\text{\hspace{1em}(4)}$$

Linear bicycle model: State space representation

Substitution of the front (3) and rear (4) tire forces yields in (1):
$$\begin{gathered} m({\dot{v}}_{lat}+\omega v_{lon})=-c_{\alpha ,r}\frac{v_{lat}-\omega l_r}{v_{lon}}+c_{\alpha ,f}\delta -c_{\alpha ,f}\frac{v_{lat}+\omega l_f}{v_{lon}} \\ {\dot{v}}_{lat}=-\frac{c_{\alpha ,r}+c_{\alpha ,f}}{m{v}_{lon}}{v}_{lat}+\frac{c_{\alpha ,r}{l}_r-c_{\alpha ,f}{l}_f}{m{v}_{lon}}\omega -v_{lon}\omega +\frac{c_{\alpha ,f}}{m}\delta \end{gathered}$$
in (2):
$$\begin{gathered} I_z\dot{\omega }=-l_r(-c_{\alpha ,r}\frac{v_{lat}-\omega l_r}{v_{lon}})+l_f(c_{\alpha ,f}\delta -c_{\alpha ,f}\frac{v_{lat}+\omega l_f}{v_{lon}}) \\ \dot{\omega }=\frac{c_{\alpha ,r}{l}_r-c_{\alpha ,f}{l}_f}{I_z{v}_{lon}}{v}_{lat}-\frac{c_{\alpha ,r}{l}_r^2-c_{\alpha ,f}{l}_f^2}{I_z{v}_{lon}}\omega +\frac{c_{\alpha ,f}}{I_z}{l}_f\delta \end{gathered}$$
This is equivalent to the following state space representation:
states: $x_1=v_{lat}$, $x_2=\psi$, $x_3=\omega$; input: $u=\delta$
$$\Rightarrow \frac{d}{dt}\left[\begin{array}{c}{x}_1\\ x_2\\ x_3\end{array}\right]=\left[\begin{array}{c}{\dot{v}}_{lat}\\ \dot{\psi }\\ \dot{\omega }\end{array}\right]=\left[\begin{array}{ccc}-\frac{c_{\alpha ,r}+c_{\alpha ,f}}{m{v}_{lon}}& 0& \frac{c_{\alpha ,r}{l}_r-c_{\alpha ,f}{l}_f}{m{v}_{lon}}-v_{lon}\\ 0& 0& 1\\ \frac{c_{\alpha ,r}{l}_r-c_{\alpha ,f}{l}_f}{I_z{v}_{lon}}& 0& -\frac{c_{\alpha ,r}{l}_r^2-c_{\alpha ,f}{l}_f^2}{I_z{v}_{lon}}\left[\begin{array}{c}{x}_1\\ x_2\\ x_3\end{array}\right]+\left[\begin{array}{c}\frac{c_{\alpha ,f}}{m}\\ 0\\ \frac{c_{\alpha ,f}}{I_z}{l}_f\end{array}\right]u\end{array}\right]$$
with $\dot{X}=v_{lon}\cos \psi -v_{lat}\sin \psi$ and $\dot{Y}=v_{lon}\sin \psi +v_{lat}\cos \psi$ the model can be augmented by the global position to a nonlinear state space model.