1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255import time
import threading
import numpy as np
import rerun as rr
from rustoracerpy import RustoracerEnv
WHEELBASE = 0.3302
STEER_FACTOR = 1.0 / 0.4189
LOOKAHEAD = 3
STEER_ALPHA = 0.35
SPEED_FAST = -0.6
SPEED_SLOW = -0.85
BRAKE_DIST = 2.0
RRT_ITERS = 200
RRT_STEP = 0.35
RRT_GOAL_BIAS = 0.30
CLEARANCE = 0.22
EDGE_RES = 0.08
OBS_RADIUS = 0.35
env = RustoracerEnv(yaml="maps/berlin.yaml", render_mode="human")
obs, info = env.reset()
N_BEAMS = env._sim.n_beams
FOV = env._sim.fov
waypoints = env.skeleton.reshape(-1, 2)
n_wps = len(waypoints)
beam_angles = np.linspace(-FOV / 2, FOV / 2, N_BEAMS)
forward_mask = np.abs(beam_angles) <= np.radians(60)
prev_steer = 0.0
rrt_path = None
obstacles: list[np.ndarray] = []
place_flag = threading.Event()
clear_flag = threading.Event()
def _input_listener():
"""Background thread: press Enter to drop an obstacle, 'c' to clear all."""
while True:
try:
line = input()
except EOFError:
break
if line.strip().lower() == "c":
clear_flag.set()
else:
place_flag.set()
threading.Thread(target=_input_listener, daemon=True).start()
print("── Press ENTER to place an obstacle at the lookahead point ──")
print("── Type 'c' + ENTER to clear all obstacles ──")
def _log_obstacles():
"""Send all obstacle circles to Rerun."""
if not obstacles:
rr.log("world/obstacles", rr.Clear(recursive=False))
return
centres_world = np.array(obstacles) # (N, 2)
centres_px = env._sim.world_to_pixels(
centres_world.ravel().astype(np.float64)
).reshape(-1, 2)
rr.log(
"world/obstacles",
rr.Points2D(
centres_px,
radii=[10.0] * len(centres_px),
colors=[[255, 50, 50, 200]] * len(centres_px),
),
)
def _hits_obstacle(pts: np.ndarray) -> bool:
"""Return True if any point in pts (N,2) is inside an obstacle."""
for o in obstacles:
if np.any(np.linalg.norm(pts - o, axis=1) < OBS_RADIUS + CLEARANCE):
return True
return False
def _edt_clear(pts):
flat = np.ascontiguousarray(pts.ravel(), dtype=np.float64)
if not bool(np.all(np.asarray(env._sim.edt_at(flat)) >= CLEARANCE)):
return False
return not _hits_obstacle(pts)
def edge_free(a, b):
d = np.linalg.norm(b - a)
if d < 1e-9:
return _edt_clear(a.reshape(1, 2))
n = max(2, int(np.ceil(d / EDGE_RES)))
pts = np.column_stack([np.linspace(a[0], b[0], n), np.linspace(a[1], b[1], n)])
return _edt_clear(pts)
def spline_goal(x, y):
pos = np.array([x, y])
nearest = int(np.argmin(np.linalg.norm(waypoints - pos, axis=1)))
for j in range(1, n_wps):
idx = (nearest + j) % n_wps
if np.linalg.norm(waypoints[idx] - pos) >= LOOKAHEAD:
return waypoints[idx]
return waypoints[(nearest + 1) % n_wps]
def pure_pursuit(x, y, theta, gx, gy):
dx, dy = gx - x, gy - y
local_y = -np.sin(theta) * dx + np.cos(theta) * dy
L2 = dx * dx + dy * dy
if L2 < 1e-12:
return 0.0
curvature = np.arctan(2.0 * local_y * WHEELBASE / L2) * STEER_FACTOR
return float(np.clip(curvature, -1, 1))
def _to_px(pt):
return env._sim.world_to_pixels(np.array(pt, dtype=np.float64)).reshape(-1, 2)
def rrt(start, goal):
nodes = [start.copy()]
parents = [-1]
lo = np.minimum(start, goal) - 2.0
hi = np.maximum(start, goal) + 2.0
for _ in range(RRT_ITERS):
sample = (
goal if np.random.random() < RRT_GOAL_BIAS else np.random.uniform(lo, hi)
)
dists = np.linalg.norm(np.array(nodes) - sample, axis=1)
ni = int(np.argmin(dists))
diff = sample - nodes[ni]
d = np.linalg.norm(diff)
if d < 1e-9:
continue
new = nodes[ni] + diff / d * min(d, RRT_STEP)
if not _edt_clear(new.reshape(1, 2)) or not edge_free(nodes[ni], new):
continue
nodes.append(new)
parents.append(ni)
if np.linalg.norm(new - goal) < RRT_STEP:
path, i = [], len(nodes) - 1
while i >= 0:
path.append(nodes[i])
i = parents[i]
path.reverse()
edges = [
np.vstack([_to_px(nodes[parents[j]]), _to_px(nodes[j])])
for j in range(1, len(nodes))
]
if edges:
rr.log("world/RRT_graph", rr.LineStrips2D(edges))
path_px = [_to_px(p) for p in path]
if len(path_px) > 1:
segments = [
np.vstack([path_px[k], path_px[k + 1]])
for k in range(len(path_px) - 1)
]
rr.log(
"world/RRT_path",
rr.LineStrips2D(segments, colors=[[0, 120, 255, 255]]),
)
return path
return None
def pick_target(path, pos):
for wp in path[1:]:
if np.linalg.norm(wp - pos) >= RRT_STEP:
return wp
return path[-1]
def forward_clearance(scans):
return float(scans[forward_mask].min())
def speed_for(clearance):
if clearance <= 0.5:
return SPEED_SLOW
if clearance < BRAKE_DIST:
t = (clearance - 0.5) / (BRAKE_DIST - 0.5)
return SPEED_SLOW + t * (SPEED_FAST - SPEED_SLOW)
return SPEED_FAST
try:
while True:
t0 = time.perf_counter()
x, y, theta, vel, *_ = info["state"][0]
pos = np.array([x, y])
scans = obs[0, :N_BEAMS]
cl = forward_clearance(scans)
goal = spline_goal(x, y)
if place_flag.is_set():
place_flag.clear()
obstacles.append(goal.copy())
print(
f" ✦ Obstacle placed at ({goal[0]:.2f}, {goal[1]:.2f}) "
f"[total: {len(obstacles)}]"
)
if clear_flag.is_set():
clear_flag.clear()
obstacles.clear()
rr.log("world/obstacles", rr.Clear(recursive=False))
print(" ✦ All obstacles cleared")
_log_obstacles()
goal_px = _to_px(goal)
rr.log(
"world/lookahead",
rr.Points2D(goal_px, radii=[5.0], colors=[[255, 200, 0, 160]]),
)
result = rrt(pos, goal)
if result and len(result) > 1:
rrt_path = result
target = pick_target(rrt_path, pos) if rrt_path else goal
raw_steer = pure_pursuit(x, y, theta, target[0], target[1])
prev_steer = STEER_ALPHA * raw_steer + (1 - STEER_ALPHA) * prev_steer
steer = float(np.clip(prev_steer, -1, 1))
obs, _, term, trunc, info = env.step(np.array([[steer, speed_for(cl)]]))
env.render()
time.sleep(max(0.0, 1 / 60 - (time.perf_counter() - t0)))
if term[0] or trunc[0]:
rrt_path, prev_steer = None, 0.0
obs, info = env.reset()
except KeyboardInterrupt:
pass
finally:
env.close()