轨迹预测

车辆正常行驶的状态可以分成三类:车道保持(Keep Lane, KL)、变道(Change Lane, CL)和转弯(Turn)。车道保持和转弯可以采用运动学模型进行预测,变道过程无法用简单的运动学模型来描述,采用特定的变道模型。

基于运动学模型的轨迹预测

基于运动学模型的轨迹预测方法有:

  1. Constant Velocity (CV):
    X k + 1 = X k + [ v t cos ⁡ ( θ ) v t sin ⁡ ( θ ) 0 0 0 0 ] T X_{k+1}=X_k+\begin{bmatrix} vt\cos(\theta) & vt\sin(\theta) & 0 & 0 & 0 & 0\end{bmatrix} ^T Xk+1=Xk+[vtcos(θ)vtsin(θ)0000]T

  2. Constant Acceleration (CA):
    X k + 1 = X k + [ ( v t + 1 / 2 a t 2 ) cos ⁡ ( θ ) ( v t + 1 / 2 a t 2 ) sin ⁡ ( θ ) 0 0 0 0 ] T X_{k+1}=X_k+\begin{bmatrix}(vt+1/2at^2)\cos(\theta) & (vt+1/2at^2)\sin(\theta) & 0 & 0 & 0 & 0\end{bmatrix} ^T Xk+1=Xk+[(vt+1/2at2)cos(θ)(vt+1/2at2)sin(θ)0000]T

  3. Constant Turn Rate and Velocity (CTRV):
    X k + 1 = X k + [ v w ( sin ⁡ ( θ + w t ) − sin ⁡ ( θ ) ) v w ( cos ⁡ ( θ ) − cos ⁡ ( θ + w t ) ) w t 0 0 0 ] T X_{k+1}=X_k+\begin{bmatrix}\frac{v}{w}(\sin(\theta +wt)-\sin(\theta)) & \frac{v}{w}(\cos(\theta)-\cos(\theta+wt)) & wt & 0 & 0 & 0\end{bmatrix} ^T Xk+1=Xk+[wv(sin(θ+wt)sin(θ))wv(cos(θ)cos(θ+wt))wt000]T

  4. Constant Turn Rate and Acceleration (CTRA):
    X k + 1 = X k + [ a [ cos ⁡ ( θ + w t ) − cos ⁡ ( θ ) ] w 2 + ( v + a t ) sin ⁡ ( θ + w t ) − v sin ⁡ ( θ ) w a [ sin ⁡ ( θ + w t ) − sin ⁡ ( θ ) ] w 2 + ( v + a t ) cos ⁡ ( θ + w t ) − v cos ⁡ ( θ ) w w t a t 0 0 ] T X_{k+1}=X_k+\begin{bmatrix} \frac{a[\cos(\theta+wt)-\cos(\theta)]}{w^2}+\frac{(v+at)\sin(\theta+wt)-v\sin(\theta)}{w} & \frac{a[\sin(\theta+wt)-\sin(\theta)]}{w^2}+\frac{(v+at)\cos(\theta+wt)-v\cos(\theta)}{w} & wt & at & 0 & 0\end{bmatrix} ^T Xk+1=Xk+[w2a[cos(θ+wt)cos(θ)]+w(v+at)sin(θ+wt)vsin(θ)w2a[sin(θ+wt)sin(θ)]+w(v+at)cos(θ+wt)vcos(θ)wtat00]T

  5. Constant Curvature and Acceleration (CCA):
    X k + 1 = X k + [ 1 c ( sin ⁡ ( θ + l c ) − sin ⁡ ( θ ) ) 1 c ( cos ⁡ ( θ ) − cos ⁡ ( θ + l c ) ) l c a t a t c 0 ] T l = v t + 1 2 a t 2 X_{k+1}=X_k+\begin{bmatrix} \frac{1}{c}(\sin(\theta+lc)-\sin(\theta)) & \frac{1}{c}(\cos(\theta)-\cos(\theta+lc)) & lc & at & atc & 0 \end{bmatrix} ^T \\ l=vt+\frac{1}{2}at^2 Xk+1=Xk+[c1(sin(θ+lc)sin(θ))c1(cos(θ)cos(θ+lc))lcatatc0]Tl=vt+21at2
    其中,状态变量为 X k = [ x , y , θ , v , w , a ] T X_k=[x,y,\theta,v,w,a]^T Xk=[x,y,θ,v,w,a]T

变道模型

采集车辆变道数据可以发现,变道过程中车辆侧向加速度的变化可以近似为一个正弦函数,如下:
a y = { a 0 + A sin ⁡ 2 π ( t − t s t a r t ) T t s t a r t ≤ t ≤ t s t a r t + T 0 其 他 a_y=\left\{\begin{array}{lc}a_0+A\sin\frac{2\pi(t-t_{start})}{T} & t_{start}\leq t \leq t_{start}+T \\0 & 其他\end{array}\right. ay={ a0+AsinT2π(ttstart)0tstartttstart+T
a y a_y ay进行两次积分,可以得到侧向位移随时间变化关系:
Y ( t ) = { A T 2 2 π + 1 2 a 0 T 2 t > t s t a r t + T 1 2 a 0 ( t − t s t a r t ) 2 + A T 2 π ( t − t s t a r t ) − A T 2 4 π 2 sin ⁡ ( 2 π T ( t − t s t a r t ) ) t s t a r t ≤ t ≤ t s t a r t + T 0 t < t s t a r t Y(t)=\left\{ \begin{array}{lcl} \frac{AT^2}{2\pi}+\frac{1}{2}a_0T^2 & & t>t_{start}+T \\ \frac{1}{2}a_0(t-t_{start})^2+\frac{AT}{2\pi}(t-t_{start})-\frac{AT^2}{4\pi^2}\sin(\frac{2\pi}{T}(t-t_{start})) & & t_{start} \leq t \leq t_{start}+T \\ 0 & & tY(t)=2πAT2+21a0T221a0(ttstart)2+2πAT(ttstart)4π2AT2sin(T2π(ttstart))0t>tstart+Ttstartttstart+Tt<tstart
纵向可近似为CA模型:
X ( t ) = v t + 1 2 a t 2 X(t)=vt+\frac{1}{2}at^2 X(t)=vt+21at2

你可能感兴趣的:(数学,深度学习)