对于车辆横向模型预测控制来说, 建立一个合适的车辆模型是非常重要的, 尤其对于高速情况下的自动驾驶, 如果我们不使用基于模型的控制(Model-free lateral Control)或者预测模型不够准确的话, 车辆对于目标航向角度会特别敏感, 车辆特别容易来回正弦抖动.
自行车运动学模型(Kinematic Bicycle Model)
运动学模型不考虑车辆的受力情况, 运动学模型只根据系统的几何关系. 并且我们假设轮胎没有滑移, 并且前轮和后轮的速度都位于车辆纵轴方向.
x˙=vcos(ψ+β)y˙=vsin(ψ+β)ψ˙=lr+lfvcos(β)tan(δf)
状态变量一共有三个z=[x,y,ψ]T, 输入变量为u=[vx,δf]
各个变量的说明如下
变量名称 |
含义 |
x |
inertial X coordinate of CoM |
y |
inertial Y coordinate of CoM |
ψ |
global heading angle |
vx |
longitudinal velocity of the vehicle at CoM |
vy |
lateral velocity of the vehicle at CoM |
v |
velocity of the vehicle at CoM, v=vx2+vy2
|
δf |
steering angle of the front wheel |
lf |
distance from the CoM to the rear axle |
lf |
distance from the CoM to the front axle |
β |
side slip angle at CoM, β=tan−1(lr+lflrtan(δf))≈vxvy
|
自行车动力学模型(Dynamic Bicycle Model)
假设车辆纵向速度变化较慢, 前轮转角不大并且使用线性轮胎模型, 我们可以使用如下动力学模型去预测车辆的状态
v˙x=wzvy+axv˙y=mCf(δf−vxvy+lfωz)+mCr(−vxvy−lrωz)−ωzvxωz˙=IzlfCf(δf−vxvy+lfωz)−IzlrCr(−vxvy−lrωz)x˙=vxcos(ψ)−vysin(ψ)y˙=vxsin(ψ)+vycos(ψ)ψ˙=ωz
一共有6个状态量z=[vx, vy, ωz, x, y, ψ]T, 输入量为u=[ax,δf]T
各个变量的说明如下
变量名称 |
含义 |
x |
inertial X coordinate of CoM |
y |
inertial Y coordinate of CoM |
ψ |
global heading angle |
vx |
longitudinal velocity of the vehicle |
δf |
steering angle of the front wheel |
lf |
distance from the CoM to the rear axle |
lf |
distance from the CoM to the front axle |
m |
vehicle mass |
ωz |
yaw rate |
Cr |
rear tire cornering stiffness |
Cf |
front tire cornering stiffness |
Iz |
yaw moment of inertia |