1 """
2 version2.0,增加环境动态
3 version1.3
4 Mobile robot motion planning sample with Dynamic Window Approach
5 结合https://blog.csdn.net/heyijia0327/article/details/44983551来看,里面含中文注释
6 符号参考《煤矿救援机器人地图构建与路径规划研究》矿大硕士论文
7 """
8
9 import math
10 import numpy as np
11 import matplotlib.pyplot as plt
12
13
14 class Config(object):
15 """
16 用来仿真的参数,
17 """
18
19 def __init__(self):
20 # robot parameter
21 self.max_speed = 1.4 # [m/s] # 最大速度
22 # self.min_speed = -0.5 # [m/s] # 最小速度,设置为可以倒车
23 self.min_speed = 0 # [m/s] # 最小速度,设置为不倒车
24 self.max_yawrate = 40.0 * math.pi / 180.0 # [rad/s] # 最大角速度
25 self.max_accel = 0.2 # [m/ss] # 最大加速度
26 self.max_dyawrate = 40.0 * math.pi / 180.0 # [rad/ss] # 最大角加速度
27 self.v_reso = 0.01 # [m/s],速度分辨率
28 self.yawrate_reso = 0.1 * math.pi / 180.0 # [rad/s],角速度分辨率
29 self.dt = 0.1 # [s] # 采样周期
30 self.predict_time = 3.0 # [s] # 向前预估三秒
31 self.to_goal_cost_gain = 1.0 # 目标代价增益
32 self.speed_cost_gain = 1.0 # 速度代价增益
33 self.robot_radius = 1.0 # [m] # 机器人半径
34
35
36 def motion(x, u, dt):
37 """
38 :param x: 位置参数,在此叫做位置空间
39 :param u: 速度和加速度,在此叫做速度空间
40 :param dt: 采样时间
41 :return:
42 """
43 # 速度更新公式比较简单,在极短时间内,车辆位移也变化较大
44 # 采用圆弧求解如何?
45 x[0] += u[0] * math.cos(x[2]) * dt # x方向位移
46 x[1] += u[0] * math.sin(x[2]) * dt # y
47 x[2] += u[1] * dt # 航向角
48 x[3] = u[0] # 速度v
49 x[4] = u[1] # 角速度w
50 # print(x)
51
52 return x
53
54
55 def calc_dynamic_window(x, config):
56 """
57 位置空间集合
58 :param x:当前位置空间,符号参考硕士论文
59 :param config:
60 :return:目前是两个速度的交集,还差一个
61 """
62
63 # 车辆能够达到的最大最小速度
64 vs = [config.min_speed, config.max_speed,
65 -config.max_yawrate, config.max_yawrate]
66
67 # 一个采样周期能够变化的最大最小速度
68 vd = [x[3] - config.max_accel * config.dt,
69 x[3] + config.max_accel * config.dt,
70 x[4] - config.max_dyawrate * config.dt,
71 x[4] + config.max_dyawrate * config.dt]
72 # print(Vs, Vd)
73
74 # 求出两个速度集合的交集
75 vr = [max(vs[0], vd[0]), min(vs[1], vd[1]),
76 max(vs[2], vd[2]), min(vs[3], vd[3])]
77
78 return vr
79
80
81 def calc_trajectory(x_init, v, w, config):
82 """
83 预测3秒内的轨迹
84 :param x_init:位置空间
85 :param v:速度
86 :param w:角速度
87 :param config:
88 :return: 每一次采样更新的轨迹,位置空间垂直堆叠
89 """
90 x = np.array(x_init)
91 trajectory = np.array(x)
92 time = 0
93 while time <= config.predict_time:
94 x = motion(x, [v, w], config.dt)
95 trajectory = np.vstack((trajectory, x)) # 垂直堆叠,vertical
96 time += config.dt
97
98 # print(trajectory)
99 return trajectory
100
101
102 def calc_to_goal_cost(trajectory, goal, config):
103 """
104 计算轨迹到目标点的代价
105 :param trajectory:轨迹搜索空间
106 :param goal:
107 :param config:
108 :return: 轨迹到目标点欧式距离
109 """
110 # calc to goal cost. It is 2D norm.
111
112 dx = goal[0] - trajectory[-1, 0]
113 dy = goal[1] - trajectory[-1, 1]
114 goal_dis = math.sqrt(dx ** 2 + dy ** 2)
115 cost = config.to_goal_cost_gain * goal_dis
116
117 return cost
118
119
120 def calc_obstacle_cost(traj, ob, config):
121 """
122 计算预测轨迹和障碍物的最小距离,dist(v,w)
123 :param traj:
124 :param ob:
125 :param config:
126 :return:
127 """
128 # calc obstacle cost inf: collision, 0:free
129
130 min_r = float("inf") # 距离初始化为无穷大
131
132 for ii in range(0, len(traj[:, 1])):
133 for i in range(len(ob[:, 0])):
134 ox = ob[i, 0]
135 oy = ob[i, 1]
136 dx = traj[ii, 0] - ox
137 dy = traj[ii, 1] - oy
138
139 r = math.sqrt(dx ** 2 + dy ** 2)
140 if r <= config.robot_radius:
141 return float("Inf") # collision
142
143 if min_r >= r:
144 min_r = r
145
146 return 1.0 / min_r # 越小越好
147
148
149 def calc_final_input(x, u, vr, config, goal, ob):
150 """
151 计算采样空间的评价函数,选择最合适的那一个作为最终输入
152 :param x:位置空间
153 :param u:速度空间
154 :param vr:速度空间交集
155 :param config:
156 :param goal:目标位置
157 :param ob:障碍物
158 :return:
159 """
160 x_init = x[:]
161 min_cost = 10000.0
162 min_u = u
163
164 best_trajectory = np.array([x])
165
166 trajectory_space = np.array([x]) # 记录搜索所有采样的轨迹,用来画图
167
168 # evaluate all trajectory with sampled input in dynamic window
169 # v,生成一系列速度,w,生成一系列角速度
170 for v in np.arange(vr[0], vr[1], config.v_reso):
171 for w in np.arange(vr[2], vr[3], config.yawrate_reso):
172
173 trajectory = calc_trajectory(x_init, v, w, config)
174
175 trajectory_space = np.vstack((trajectory_space, trajectory))
176
177 # calc cost
178 to_goal_cost = calc_to_goal_cost(trajectory, goal, config)
179 speed_cost = config.speed_cost_gain * (config.max_speed - trajectory[-1, 3])
180 ob_cost = calc_obstacle_cost(trajectory, ob, config)
181 # print(ob_cost)
182
183 # 评价函数多种多样,看自己选择
184 # 本文构造的是越小越好
185 final_cost = to_goal_cost + speed_cost + ob_cost
186
187 # search minimum trajectory
188 if min_cost >= final_cost:
189 min_cost = final_cost
190 min_u = [v, w]
191 best_trajectory = trajectory
192
193 # print(min_u)
194 # input()
195
196 return min_u, best_trajectory, trajectory_space
197
198
199 def dwa_control(x, u, config, goal, ob):
200 """
201 调用前面的几个函数,生成最合适的速度空间和轨迹搜索空间
202 :param x:
203 :param u:
204 :param config:
205 :param goal:
206 :param ob:
207 :return:
208 """
209 # Dynamic Window control
210
211 vr = calc_dynamic_window(x, config)
212
213 u, trajectory, trajectory_space = calc_final_input(x, u, vr, config, goal, ob)
214
215 return u, trajectory, trajectory_space
216
217
218 def plot_arrow(x, y, yaw, length=0.5, width=0.1):
219 """
220 arrow函数绘制箭头,表示搜索过程中选择的航向角
221 :param x:
222 :param y:
223 :param yaw:航向角
224 :param length:
225 :param width:参数值为浮点数,代表箭头尾部的宽度,默认值为0.001
226 :return:
227 length_includes_head:代表箭头整体长度是否包含箭头头部的长度,默认值为False
228 head_width:代表箭头头部的宽度,默认值为3*width,即尾部宽度的3倍
229 head_length:代表箭头头部的长度度,默认值为1.5*head_width,即头部宽度的1.5倍
230 shape:参数值为'full'、'left'、'right',表示箭头的形状,默认值为'full'
231 overhang:代表箭头头部三角形底边与箭头尾部直接的夹角关系,通过该参数可改变箭头的形状。
232 默认值为0,即头部为三角形,当该值小于0时,头部为菱形,当值大于0时,头部为鱼尾状
233 """
234 plt.arrow(x, y, length * math.cos(yaw), length * math.sin(yaw),
235 head_length=1.5 * width, head_width=width)
236 plt.plot(x, y)
237
238
239 def dynamic_obstacle():
240 """
241 生成多个障碍物,但是不能生成在起点和终点
242 :return:
243 """
244 obstacle = np.random.randint(1, 10, size=(10, 2))
245 return obstacle
246
247
248 def main():
249 """
250 主函数
251 :return:
252 """
253 # print(__file__ + " start!!")
254 # 初始化位置空间
255 x = np.array([0.0, 0.0, math.pi / 2.0, 0.0, 0.0])
256 goal = np.array([10, 10])
257
258 u = np.array([0.0, 0.0])
259 config = Config()
260 trajectory = np.array(x)
261 # obstacle_time = 0
262
263 while True:
264
265 ob = dynamic_obstacle() # 障碍物初始化
266
267 u, best_trajectory, trajectory_space = dwa_control(x, u, config, goal, ob)
268
269 # 据前面计算的结果使曲线前进
270 x = motion(x, u, config.dt)
271 # print(x)
272
273 trajectory = np.vstack((trajectory, x)) # store state history
274
275 # 画出每次前进的结果
276 draw_dynamic_search(best_trajectory, x, goal, ob, trajectory_space)
277
278 # check goal,小于机器人半径,则搜索结束
279 if math.sqrt((x[0] - goal[0]) ** 2 + (x[1] - goal[1]) ** 2) <= config.robot_radius:
280 print("Goal!!")
281
282 break
283
284 # obstacle_time += config.dt
285
286 print("Done")
287
288
289 def draw_dynamic_search(best_trajectory, x, goal, ob, trajectory_space):
290 """
291 画出动态搜索过程图
292 :return:
293 """
294 plt.cla() # 清除上次绘制图像
295 plt.plot(best_trajectory[:, 0], best_trajectory[:, 1], "-g")
296
297 plt.plot(trajectory_space[:, 0], trajectory_space[:, 1], '-g')
298
299 plt.plot(x[0], x[1], "xr")
300 plt.plot(0, 0, "og")
301 plt.plot(goal[0], goal[1], "ro")
302 plt.plot(ob[:, 0], ob[:, 1], "bo")
303 plot_arrow(x[0], x[1], x[2])
304 plt.axis("equal")
305 plt.grid(True)
306 plt.pause(0.0001)
307
308
309 def draw_path(trajectory, goal, ob, x):
310 """
311 画图函数
312 :return:
313 """
314 plt.cla() # 清除上次绘制图像
315
316 plt.plot(x[0], x[1], "xr")
317 plt.plot(0, 0, "og")
318 plt.plot(goal[0], goal[1], "ro")
319 plt.plot(ob[:, 0], ob[:, 1], "bs")
320 plot_arrow(x[0], x[1], x[2])
321 plt.axis("equal")
322 plt.grid(True)
323 plt.plot(trajectory[:, 0], trajectory[:, 1], 'r')
324 plt.show()
325
326
327 if __name__ == '__main__':
328 main()