|
| 1 | +""" |
| 2 | +
|
| 3 | +Path planning Sample Code with Randomized Rapidly-Exploring Random |
| 4 | +Trees with sobol low discrepancy sampler(RRTSobol). |
| 5 | +Sobol wiki https://en.wikipedia.org/wiki/Sobol_sequence |
| 6 | +
|
| 7 | +The goal of low discrepancy samplers is to generate a sequence of points that |
| 8 | +optimizes a criterion called dispersion. Intuitively, the idea is to place |
| 9 | +samples to cover the exploration space in a way that makes the largest |
| 10 | +uncovered area be as small as possible. This generalizes of the idea of grid |
| 11 | +resolution. For a grid, the resolution may be selected by defining the step |
| 12 | +size for each axis. As the step size is decreased, the resolution increases. |
| 13 | +If a grid-based motion planning algorithm can increase the resolution |
| 14 | +arbitrarily, it becomes resolution complete. Dispersion can be considered as a |
| 15 | +powerful generalization of the notion of resolution. |
| 16 | +
|
| 17 | +Taken from |
| 18 | +LaValle, Steven M. Planning algorithms. Cambridge university press, 2006. |
| 19 | +
|
| 20 | +
|
| 21 | +authors: |
| 22 | + First implementation AtsushiSakai(@Atsushi_twi) |
| 23 | + Sobol sampler Rafael A. |
| 24 | + |
| 25 | +
|
| 26 | +
|
| 27 | +""" |
| 28 | + |
| 29 | +import math |
| 30 | +import random |
| 31 | +from sobol import sobol_quasirand |
| 32 | +import sys |
| 33 | + |
| 34 | +import matplotlib.pyplot as plt |
| 35 | +import numpy as np |
| 36 | + |
| 37 | +show_animation = True |
| 38 | + |
| 39 | + |
| 40 | +class RRTSobol: |
| 41 | + """ |
| 42 | + Class for RRTSobol planning |
| 43 | + """ |
| 44 | + |
| 45 | + class Node: |
| 46 | + """ |
| 47 | + RRTSobol Node |
| 48 | + """ |
| 49 | + |
| 50 | + def __init__(self, x, y): |
| 51 | + self.x = x |
| 52 | + self.y = y |
| 53 | + self.path_x = [] |
| 54 | + self.path_y = [] |
| 55 | + self.parent = None |
| 56 | + |
| 57 | + def __init__(self, |
| 58 | + start, |
| 59 | + goal, |
| 60 | + obstacle_list, |
| 61 | + rand_area, |
| 62 | + expand_dis=3.0, |
| 63 | + path_resolution=0.5, |
| 64 | + goal_sample_rate=5, |
| 65 | + max_iter=500): |
| 66 | + """ |
| 67 | + Setting Parameter |
| 68 | +
|
| 69 | + start:Start Position [x,y] |
| 70 | + goal:Goal Position [x,y] |
| 71 | + obstacle_list:obstacle Positions [[x,y,size],...] |
| 72 | + randArea:Random Sampling Area [min,max] |
| 73 | +
|
| 74 | + """ |
| 75 | + self.start = self.Node(start[0], start[1]) |
| 76 | + self.end = self.Node(goal[0], goal[1]) |
| 77 | + self.min_rand = rand_area[0] |
| 78 | + self.max_rand = rand_area[1] |
| 79 | + self.expand_dis = expand_dis |
| 80 | + self.path_resolution = path_resolution |
| 81 | + self.goal_sample_rate = goal_sample_rate |
| 82 | + self.max_iter = max_iter |
| 83 | + self.obstacle_list = obstacle_list |
| 84 | + self.node_list = [] |
| 85 | + self.sobol_inter_ = 0 |
| 86 | + |
| 87 | + def planning(self, animation=True): |
| 88 | + """ |
| 89 | + rrt path planning |
| 90 | +
|
| 91 | + animation: flag for animation on or off |
| 92 | + """ |
| 93 | + |
| 94 | + self.node_list = [self.start] |
| 95 | + for i in range(self.max_iter): |
| 96 | + rnd_node = self.get_random_node() |
| 97 | + nearest_ind = self.get_nearest_node_index(self.node_list, rnd_node) |
| 98 | + nearest_node = self.node_list[nearest_ind] |
| 99 | + |
| 100 | + new_node = self.steer(nearest_node, rnd_node, self.expand_dis) |
| 101 | + |
| 102 | + if self.check_collision(new_node, self.obstacle_list): |
| 103 | + self.node_list.append(new_node) |
| 104 | + |
| 105 | + if animation and i % 5 == 0: |
| 106 | + self.draw_graph(rnd_node) |
| 107 | + |
| 108 | + if self.calc_dist_to_goal(self.node_list[-1].x, |
| 109 | + self.node_list[-1].y) <= self.expand_dis: |
| 110 | + final_node = self.steer(self.node_list[-1], self.end, |
| 111 | + self.expand_dis) |
| 112 | + if self.check_collision(final_node, self.obstacle_list): |
| 113 | + return self.generate_final_course(len(self.node_list) - 1) |
| 114 | + |
| 115 | + if animation and i % 5: |
| 116 | + self.draw_graph(rnd_node) |
| 117 | + |
| 118 | + return None # cannot find path |
| 119 | + |
| 120 | + def steer(self, from_node, to_node, extend_length=float("inf")): |
| 121 | + |
| 122 | + new_node = self.Node(from_node.x, from_node.y) |
| 123 | + d, theta = self.calc_distance_and_angle(new_node, to_node) |
| 124 | + |
| 125 | + new_node.path_x = [new_node.x] |
| 126 | + new_node.path_y = [new_node.y] |
| 127 | + |
| 128 | + if extend_length > d: |
| 129 | + extend_length = d |
| 130 | + |
| 131 | + n_expand = math.floor(extend_length / self.path_resolution) |
| 132 | + |
| 133 | + for _ in range(n_expand): |
| 134 | + new_node.x += self.path_resolution * math.cos(theta) |
| 135 | + new_node.y += self.path_resolution * math.sin(theta) |
| 136 | + new_node.path_x.append(new_node.x) |
| 137 | + new_node.path_y.append(new_node.y) |
| 138 | + |
| 139 | + d, _ = self.calc_distance_and_angle(new_node, to_node) |
| 140 | + if d <= self.path_resolution: |
| 141 | + new_node.path_x.append(to_node.x) |
| 142 | + new_node.path_y.append(to_node.y) |
| 143 | + new_node.x = to_node.x |
| 144 | + new_node.y = to_node.y |
| 145 | + |
| 146 | + new_node.parent = from_node |
| 147 | + |
| 148 | + return new_node |
| 149 | + |
| 150 | + def generate_final_course(self, goal_ind): |
| 151 | + path = [[self.end.x, self.end.y]] |
| 152 | + node = self.node_list[goal_ind] |
| 153 | + while node.parent is not None: |
| 154 | + path.append([node.x, node.y]) |
| 155 | + node = node.parent |
| 156 | + path.append([node.x, node.y]) |
| 157 | + |
| 158 | + return path |
| 159 | + |
| 160 | + def calc_dist_to_goal(self, x, y): |
| 161 | + dx = x - self.end.x |
| 162 | + dy = y - self.end.y |
| 163 | + return math.hypot(dx, dy) |
| 164 | + |
| 165 | + def get_random_node(self): |
| 166 | + if random.randint(0, 100) > self.goal_sample_rate: |
| 167 | + rand_coordinates, n = sobol_quasirand(2, self.sobol_inter_) |
| 168 | + |
| 169 | + rand_coordinates = self.min_rand + \ |
| 170 | + rand_coordinates*(self.max_rand - self.min_rand) |
| 171 | + self.sobol_inter_ = n |
| 172 | + rnd = self.Node(*rand_coordinates) |
| 173 | + |
| 174 | + else: # goal point sampling |
| 175 | + rnd = self.Node(self.end.x, self.end.y) |
| 176 | + return rnd |
| 177 | + |
| 178 | + def draw_graph(self, rnd=None): |
| 179 | + plt.clf() |
| 180 | + # for stopping simulation with the esc key. |
| 181 | + plt.gcf().canvas.mpl_connect( |
| 182 | + 'key_release_event', |
| 183 | + lambda event: [sys.exit(0) if event.key == 'escape' else None]) |
| 184 | + if rnd is not None: |
| 185 | + plt.plot(rnd.x, rnd.y, "^k") |
| 186 | + for node in self.node_list: |
| 187 | + if node.parent: |
| 188 | + plt.plot(node.path_x, node.path_y, "-g") |
| 189 | + |
| 190 | + for (ox, oy, size) in self.obstacle_list: |
| 191 | + self.plot_circle(ox, oy, size) |
| 192 | + |
| 193 | + plt.plot(self.start.x, self.start.y, "xr") |
| 194 | + plt.plot(self.end.x, self.end.y, "xr") |
| 195 | + plt.axis("equal") |
| 196 | + plt.axis([-2, 15, -2, 15]) |
| 197 | + plt.grid(True) |
| 198 | + plt.pause(0.01) |
| 199 | + |
| 200 | + @staticmethod |
| 201 | + def plot_circle(x, y, size, color="-b"): # pragma: no cover |
| 202 | + deg = list(range(0, 360, 5)) |
| 203 | + deg.append(0) |
| 204 | + xl = [x + size * math.cos(np.deg2rad(d)) for d in deg] |
| 205 | + yl = [y + size * math.sin(np.deg2rad(d)) for d in deg] |
| 206 | + plt.plot(xl, yl, color) |
| 207 | + |
| 208 | + @staticmethod |
| 209 | + def get_nearest_node_index(node_list, rnd_node): |
| 210 | + dlist = [(node.x - rnd_node.x)**2 + (node.y - rnd_node.y)**2 |
| 211 | + for node in node_list] |
| 212 | + minind = dlist.index(min(dlist)) |
| 213 | + |
| 214 | + return minind |
| 215 | + |
| 216 | + @staticmethod |
| 217 | + def check_collision(node, obstacle_list): |
| 218 | + |
| 219 | + if node is None: |
| 220 | + return False |
| 221 | + |
| 222 | + for (ox, oy, size) in obstacle_list: |
| 223 | + dx_list = [ox - x for x in node.path_x] |
| 224 | + dy_list = [oy - y for y in node.path_y] |
| 225 | + d_list = [dx * dx + dy * dy for (dx, dy) in zip(dx_list, dy_list)] |
| 226 | + |
| 227 | + if min(d_list) <= size**2: |
| 228 | + return False # collision |
| 229 | + |
| 230 | + return True # safe |
| 231 | + |
| 232 | + @staticmethod |
| 233 | + def calc_distance_and_angle(from_node, to_node): |
| 234 | + dx = to_node.x - from_node.x |
| 235 | + dy = to_node.y - from_node.y |
| 236 | + d = math.hypot(dx, dy) |
| 237 | + theta = math.atan2(dy, dx) |
| 238 | + return d, theta |
| 239 | + |
| 240 | + |
| 241 | +def main(gx=6.0, gy=10.0): |
| 242 | + print("start " + __file__) |
| 243 | + |
| 244 | + # ====Search Path with RRTSobol==== |
| 245 | + obstacle_list = [(5, 5, 1), (3, 6, 2), (3, 8, 2), (3, 10, 2), (7, 5, 2), |
| 246 | + (9, 5, 2), (8, 10, 1)] # [x, y, radius] |
| 247 | + # Set Initial parameters |
| 248 | + rrt = RRTSobol( |
| 249 | + start=[0, 0], |
| 250 | + goal=[gx, gy], |
| 251 | + rand_area=[-2, 15], |
| 252 | + obstacle_list=obstacle_list) |
| 253 | + path = rrt.planning(animation=show_animation) |
| 254 | + |
| 255 | + if path is None: |
| 256 | + print("Cannot find path") |
| 257 | + else: |
| 258 | + print("found path!!") |
| 259 | + |
| 260 | + # Draw final path |
| 261 | + if show_animation: |
| 262 | + rrt.draw_graph() |
| 263 | + plt.plot([x for (x, y) in path], [y for (x, y) in path], '-r') |
| 264 | + plt.grid(True) |
| 265 | + plt.pause(0.01) # Need for Mac |
| 266 | + plt.show() |
| 267 | + |
| 268 | + |
| 269 | +if __name__ == '__main__': |
| 270 | + main() |
0 commit comments