Skip to content

Commit 9274aac

Browse files
authored
Merge pull request AtsushiSakai#303 from Chachay/MPC
[Proposal] Reduce internal dependancy
2 parents 55478fa + b6e89c6 commit 9274aac

File tree

1 file changed

+101
-107
lines changed

1 file changed

+101
-107
lines changed

PathTracking/rear_wheel_feedback/rear_wheel_feedback.py

Lines changed: 101 additions & 107 deletions
Original file line numberDiff line numberDiff line change
@@ -8,14 +8,9 @@
88
import matplotlib.pyplot as plt
99
import math
1010
import numpy as np
11-
import sys
12-
sys.path.append("../../PathPlanning/CubicSpline/")
13-
14-
try:
15-
import cubic_spline_planner
16-
except:
17-
raise
1811

12+
from scipy import interpolate
13+
from scipy import optimize
1914

2015
Kp = 1.0 # speed propotional gain
2116
# steering control parameter
@@ -26,34 +21,80 @@
2621
L = 2.9 # [m]
2722

2823
show_animation = True
29-
# show_animation = False
30-
3124

3225
class State:
33-
34-
def __init__(self, x=0.0, y=0.0, yaw=0.0, v=0.0):
26+
def __init__(self, x=0.0, y=0.0, yaw=0.0, v=0.0, direction=1):
3527
self.x = x
3628
self.y = y
3729
self.yaw = yaw
3830
self.v = v
39-
40-
41-
def update(state, a, delta):
42-
43-
state.x = state.x + state.v * math.cos(state.yaw) * dt
44-
state.y = state.y + state.v * math.sin(state.yaw) * dt
45-
state.yaw = state.yaw + state.v / L * math.tan(delta) * dt
46-
state.v = state.v + a * dt
47-
48-
return state
49-
50-
51-
def PIDControl(target, current):
31+
self.direction = direction
32+
33+
def update(self, a, delta, dt):
34+
self.x = self.x + self.v * math.cos(self.yaw) * dt
35+
self.y = self.y + self.v * math.sin(self.yaw) * dt
36+
self.yaw = self.yaw + self.v / L * math.tan(delta) * dt
37+
self.v = self.v + a * dt
38+
39+
class CubicSplinePath:
40+
def __init__(self, x, y):
41+
x, y = map(np.asarray, (x, y))
42+
s = np.append([0],(np.cumsum(np.diff(x)**2) + np.cumsum(np.diff(y)**2))**0.5)
43+
44+
self.X = interpolate.CubicSpline(s, x)
45+
self.Y = interpolate.CubicSpline(s, y)
46+
47+
self.dX = self.X.derivative(1)
48+
self.ddX = self.X.derivative(2)
49+
50+
self.dY = self.Y.derivative(1)
51+
self.ddY = self.Y.derivative(2)
52+
53+
self.length = s[-1]
54+
55+
def calc_yaw(self, s):
56+
dx, dy = self.dX(s), self.dY(s)
57+
return np.arctan2(dy, dx)
58+
59+
def calc_curvature(self, s):
60+
dx, dy = self.dX(s), self.dY(s)
61+
ddx, ddy = self.ddX(s), self.ddY(s)
62+
return (ddy * dx - ddx * dy) / ((dx ** 2 + dy ** 2)**(3 / 2))
63+
64+
def __find_nearest_point(self, s0, x, y):
65+
def calc_distance(_s, *args):
66+
_x, _y= self.X(_s), self.Y(_s)
67+
return (_x - args[0])**2 + (_y - args[1])**2
68+
69+
def calc_distance_jacobian(_s, *args):
70+
_x, _y = self.X(_s), self.Y(_s)
71+
_dx, _dy = self.dX(_s), self.dY(_s)
72+
return 2*_dx*(_x - args[0])+2*_dy*(_y-args[1])
73+
74+
minimum = optimize.fmin_cg(calc_distance, s0, calc_distance_jacobian, args=(x, y), full_output=True, disp=False)
75+
return minimum
76+
77+
def calc_track_error(self, x, y, s0):
78+
ret = self.__find_nearest_point(s0, x, y)
79+
80+
s = ret[0][0]
81+
e = ret[1]
82+
83+
k = self.calc_curvature(s)
84+
yaw = self.calc_yaw(s)
85+
86+
dxl = self.X(s) - x
87+
dyl = self.Y(s) - y
88+
angle = pi_2_pi(yaw - math.atan2(dyl, dxl))
89+
if angle < 0:
90+
e*= -1
91+
92+
return e, k, yaw, s
93+
94+
def pid_control(target, current):
5295
a = Kp * (target - current)
53-
5496
return a
5597

56-
5798
def pi_2_pi(angle):
5899
while(angle > math.pi):
59100
angle = angle - 2.0 * math.pi
@@ -63,53 +104,24 @@ def pi_2_pi(angle):
63104

64105
return angle
65106

66-
67-
def rear_wheel_feedback_control(state, cx, cy, cyaw, ck, preind):
68-
ind, e = calc_nearest_index(state, cx, cy, cyaw)
69-
70-
k = ck[ind]
107+
def rear_wheel_feedback_control(state, e, k, yaw_ref):
71108
v = state.v
72-
th_e = pi_2_pi(state.yaw - cyaw[ind])
109+
th_e = pi_2_pi(state.yaw - yaw_ref)
73110

74111
omega = v * k * math.cos(th_e) / (1.0 - k * e) - \
75112
KTH * abs(v) * th_e - KE * v * math.sin(th_e) * e / th_e
76113

77114
if th_e == 0.0 or omega == 0.0:
78-
return 0.0, ind
115+
return 0.0
79116

80117
delta = math.atan2(L * omega / v, 1.0)
81-
# print(k, v, e, th_e, omega, delta)
82-
83-
return delta, ind
84-
85-
86-
def calc_nearest_index(state, cx, cy, cyaw):
87-
dx = [state.x - icx for icx in cx]
88-
dy = [state.y - icy for icy in cy]
89-
90-
d = [idx ** 2 + idy ** 2 for (idx, idy) in zip(dx, dy)]
91-
92-
mind = min(d)
93118

94-
ind = d.index(mind)
119+
return delta
95120

96-
mind = math.sqrt(mind)
97-
98-
dxl = cx[ind] - state.x
99-
dyl = cy[ind] - state.y
100-
101-
angle = pi_2_pi(cyaw[ind] - math.atan2(dyl, dxl))
102-
if angle < 0:
103-
mind *= -1
104-
105-
return ind, mind
106-
107-
108-
def closed_loop_prediction(cx, cy, cyaw, ck, speed_profile, goal):
109121

122+
def simulate(path_ref, goal):
110123
T = 500.0 # max simulation time
111124
goal_dis = 0.3
112-
stop_speed = 0.05
113125

114126
state = State(x=-0.0, y=-0.0, yaw=0.0, v=0.0)
115127

@@ -120,16 +132,17 @@ def closed_loop_prediction(cx, cy, cyaw, ck, speed_profile, goal):
120132
v = [state.v]
121133
t = [0.0]
122134
goal_flag = False
123-
target_ind = calc_nearest_index(state, cx, cy, cyaw)
135+
136+
s = np.arange(0, path_ref.length, 0.1)
137+
e, k, yaw_ref, s0 = path_ref.calc_track_error(state.x, state.y, 0.0)
124138

125139
while T >= time:
126-
di, target_ind = rear_wheel_feedback_control(
127-
state, cx, cy, cyaw, ck, target_ind)
128-
ai = PIDControl(speed_profile[target_ind], state.v)
129-
state = update(state, ai, di)
140+
e, k, yaw_ref, s0 = path_ref.calc_track_error(state.x, state.y, s0)
141+
di = rear_wheel_feedback_control(state, e, k, yaw_ref)
130142

131-
if abs(state.v) <= stop_speed:
132-
target_ind += 1
143+
speed_ref = calc_target_speed(state, yaw_ref)
144+
ai = pid_control(speed_ref, state.v)
145+
state.update(ai, di, dt)
133146

134147
time = time + dt
135148

@@ -147,64 +160,46 @@ def closed_loop_prediction(cx, cy, cyaw, ck, speed_profile, goal):
147160
v.append(state.v)
148161
t.append(time)
149162

150-
if target_ind % 1 == 0 and show_animation:
163+
if show_animation:
151164
plt.cla()
152165
# for stopping simulation with the esc key.
153166
plt.gcf().canvas.mpl_connect('key_release_event',
154167
lambda event: [exit(0) if event.key == 'escape' else None])
155-
plt.plot(cx, cy, "-r", label="course")
168+
plt.plot(path_ref.X(s), path_ref.Y(s), "-r", label="course")
156169
plt.plot(x, y, "ob", label="trajectory")
157-
plt.plot(cx[target_ind], cy[target_ind], "xg", label="target")
170+
plt.plot(path_ref.X(s0), path_ref.Y(s0), "xg", label="target")
158171
plt.axis("equal")
159172
plt.grid(True)
160-
plt.title("speed[km/h]:" + str(round(state.v * 3.6, 2)) +
161-
",target index:" + str(target_ind))
173+
plt.title("speed[km/h]:{:.2f}, target s-param:{:.2f}".format(round(state.v * 3.6, 2), s0))
162174
plt.pause(0.0001)
163175

164176
return t, x, y, yaw, v, goal_flag
165177

178+
def calc_target_speed(state, yaw_ref):
179+
target_speed = 10.0 / 3.6
166180

167-
def calc_speed_profile(cx, cy, cyaw, target_speed):
168-
169-
speed_profile = [target_speed] * len(cx)
170-
171-
direction = 1.0
172-
173-
# Set stop point
174-
for i in range(len(cx) - 1):
175-
dyaw = cyaw[i + 1] - cyaw[i]
176-
switch = math.pi / 4.0 <= dyaw < math.pi / 2.0
177-
178-
if switch:
179-
direction *= -1
180-
181-
if direction != 1.0:
182-
speed_profile[i] = - target_speed
183-
else:
184-
speed_profile[i] = target_speed
185-
186-
if switch:
187-
speed_profile[i] = 0.0
188-
189-
speed_profile[-1] = 0.0
181+
dyaw = yaw_ref - state.yaw
182+
switch = math.pi / 4.0 <= dyaw < math.pi / 2.0
190183

191-
return speed_profile
184+
if switch:
185+
state.direction *= -1
186+
return 0.0
187+
188+
if state.direction != 1:
189+
return -target_speed
192190

191+
return target_speed
193192

194193
def main():
195194
print("rear wheel feedback tracking start!!")
196195
ax = [0.0, 6.0, 12.5, 5.0, 7.5, 3.0, -1.0]
197196
ay = [0.0, 0.0, 5.0, 6.5, 3.0, 5.0, -2.0]
198197
goal = [ax[-1], ay[-1]]
199198

200-
cx, cy, cyaw, ck, s = cubic_spline_planner.calc_spline_course(
201-
ax, ay, ds=0.1)
202-
target_speed = 10.0 / 3.6
203-
204-
sp = calc_speed_profile(cx, cy, cyaw, target_speed)
199+
reference_path = CubicSplinePath(ax, ay)
200+
s = np.arange(0, reference_path.length, 0.1)
205201

206-
t, x, y, yaw, v, goal_flag = closed_loop_prediction(
207-
cx, cy, cyaw, ck, sp, goal)
202+
t, x, y, yaw, v, goal_flag = simulate(reference_path, goal)
208203

209204
# Test
210205
assert goal_flag, "Cannot goal"
@@ -213,7 +208,7 @@ def main():
213208
plt.close()
214209
plt.subplots(1)
215210
plt.plot(ax, ay, "xb", label="input")
216-
plt.plot(cx, cy, "-r", label="spline")
211+
plt.plot(reference_path.X(s), reference_path.Y(s), "-r", label="spline")
217212
plt.plot(x, y, "-g", label="tracking")
218213
plt.grid(True)
219214
plt.axis("equal")
@@ -222,21 +217,20 @@ def main():
222217
plt.legend()
223218

224219
plt.subplots(1)
225-
plt.plot(s, [np.rad2deg(iyaw) for iyaw in cyaw], "-r", label="yaw")
220+
plt.plot(s, np.rad2deg(reference_path.calc_yaw(s)), "-r", label="yaw")
226221
plt.grid(True)
227222
plt.legend()
228223
plt.xlabel("line length[m]")
229224
plt.ylabel("yaw angle[deg]")
230225

231226
plt.subplots(1)
232-
plt.plot(s, ck, "-r", label="curvature")
227+
plt.plot(s, reference_path.calc_curvature(s), "-r", label="curvature")
233228
plt.grid(True)
234229
plt.legend()
235230
plt.xlabel("line length[m]")
236231
plt.ylabel("curvature [1/m]")
237232

238233
plt.show()
239234

240-
241235
if __name__ == '__main__':
242236
main()

0 commit comments

Comments
 (0)