File size: 5,618 Bytes
b35f113 | 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 | """
Lane Change Analysis — Automatum Data Highway Area Pools (Multi-Drone)
Analyzes lane change behavior across all vehicles in a recording.
Usage:
python 01_lane_changes.py <path_to_recording_folder>
Example:
python 01_lane_changes.py ../Sample_Data/Pool-Highway-A9-OffenbauRastanlage_c025-c02598c7-fa60-4302-8cee-12a05f4e0db6
Note:
The exports of this release do not contain the per-sample lane assignment
(`lane_id_vec`). If it is missing, the script estimates lanes from the
vehicle's lateral position across the road: per driving direction a local road
axis (median lateral position in 20 m slices) is computed, the lateral offsets are
clustered into lanes (histogram peaks), and a lane change is counted when
a vehicle stays at least 1 s on a new lane.
"""
import sys
import os
import numpy as np
from openautomatumdronedata.dataset import droneDataset
def _road_axis(dynWorld):
"""Principal axis of all positions (unit vector) and the centroid."""
pts = np.vstack([np.column_stack([o.x_vec, o.y_vec]) for o in dynWorld.dynamicObjects.values()
if len(o.x_vec) > 1])
c = pts.mean(0)
_, _, vt = np.linalg.svd(pts[::50] - c, full_matrices=False)
return c, vt[0]
def estimate_lane_ids(dynWorld, lane_min_gap_m=2.5, min_dwell_s=1.0):
"""Estimated lane index per sample for every object (dict UUID -> np.array)."""
c, ax = _road_axis(dynWorld)
nrm = np.array([-ax[1], ax[0]])
objs = [o for o in dynWorld.dynamicObjects.values() if len(o.x_vec) > 1]
s, d, direction = {}, {}, {}
for o in objs:
p = np.column_stack([o.x_vec, o.y_vec]) - c
s[o.UUID], d[o.UUID] = p @ ax, p @ nrm
direction[o.UUID] = 1 if s[o.UUID][-1] >= s[o.UUID][0] else -1
lane_ids = {}
dt = dynWorld.delta_t
for dirn in (1, -1):
ids = [u for u in s if direction[u] == dirn]
if not ids:
continue
S = np.concatenate([s[u] for u in ids]); D = np.concatenate([d[u] for u in ids])
# local road axis of this direction: median lateral position in 20 m slices along the road
edges_s = np.arange(S.min(), S.max() + 20.0, 20.0)
idx = np.clip(np.digitize(S, edges_s) - 1, 0, len(edges_s) - 2)
med = np.array([np.median(D[idx == k]) if np.any(idx == k) else np.nan for k in range(len(edges_s) - 1)])
mid = 0.5 * (edges_s[1:] + edges_s[:-1]); ok = ~np.isnan(med)
res = {u: d[u] - np.interp(s[u], mid[ok], med[ok]) for u in ids}
R = np.concatenate(list(res.values()))
hist, edges = np.histogram(R, bins=np.arange(R.min() - 1, R.max() + 1, 0.25))
hist = np.convolve(hist, np.ones(5) / 5, mode="same")
centers = 0.5 * (edges[1:] + edges[:-1])
peaks = [i for i in range(1, len(hist) - 1)
if hist[i] >= hist[i - 1] and hist[i] > hist[i + 1] and hist[i] > 0.02 * hist.max()]
lanes = []
for i in sorted(peaks, key=lambda i: -hist[i]):
if all(abs(centers[i] - l) > lane_min_gap_m for l in lanes):
lanes.append(centers[i])
lanes = np.sort(np.array(lanes))
min_len = max(1, int(round(min_dwell_s / dt)))
for u in ids:
raw = np.argmin(np.abs(res[u][:, None] - lanes[None, :]), axis=1)
# hysteresis: only accept a new lane after min_dwell_s on it
out = raw.copy(); cur = raw[0]; run_start = 0
for k in range(1, len(raw)):
if raw[k] != raw[k - 1]:
run_start = k
if raw[k] != cur and k - run_start + 1 >= min_len:
cur = raw[k]
out[k] = cur
lane_ids[u] = out
return lane_ids
def analyze_lane_changes(dataset_path):
print(f"Loading dataset from: {dataset_path}")
dataset = droneDataset(dataset_path)
dynWorld = dataset.dynWorld
print(f"Vehicles found: {len(dynWorld)}")
have_lane_ids = any(getattr(o, "lane_id_vec", None) is not None and len(o.lane_id_vec) > 0
for o in dynWorld.dynamicObjects.values())
estimated = None
if not have_lane_ids:
print("lane_id_vec is not part of this export -> estimating lanes from the lateral position")
estimated = estimate_lane_ids(dynWorld)
lane_change_counts = []
for dynObj in dynWorld.dynamicObjects.values():
lane_ids = dynObj.lane_id_vec if have_lane_ids else estimated.get(dynObj.UUID)
if lane_ids is None or len(lane_ids) == 0:
continue
changes = 0
current_lane = lane_ids[0]
for lane in lane_ids[1:]:
if lane != current_lane:
changes += 1
current_lane = lane
lane_change_counts.append({
"uuid": dynObj.UUID,
"type": dynObj.type,
"changes": changes,
})
sorted_by_changes = sorted(lane_change_counts, key=lambda x: x["changes"], reverse=True)
n_changing = sum(1 for x in lane_change_counts if x["changes"] > 0)
total = sum(x["changes"] for x in lane_change_counts)
print(f"Lane changes: {total} in total, {n_changing} of {len(lane_change_counts)} vehicles change lane at least once")
print("\n--- Top 10 vehicles with most lane changes ---")
for idx, item in enumerate(sorted_by_changes[:10]):
print(f"{idx+1}. Vehicle {item['uuid']} ({item['type']}): {item['changes']} lane changes")
if __name__ == "__main__":
if len(sys.argv) < 2:
print("Usage: python 01_lane_changes.py <path_to_recording_folder>")
else:
analyze_lane_changes(sys.argv[1])
|