Add avoidance and intrusion number computation

[schroedtert]

No description provided.

[schroedtert]

@chraibi can you take over and add the documentation for the functions?

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 1 comment.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

compute_individual_distances builds a full Cartesian product of pedestrians per frame via merge, which is O(n²) rows and can become prohibitively slow/memory-intensive for typical crowd sizes. Consider adding an optional max-distance/cutoff parameter (or a separate compute_individual_distances_within_radius) and using a spatial index (e.g., Shapely STRtree.query/query_nearest) to only materialize pairs that can actually interact, rather than all pairs.

Suggested change

	def compute_individual_distances(*, traj_data: TrajectoryData) -> pd.DataFrame:
	"""Compute pairwise distances between all pedestrians per frame.
	Args:
	traj_data: trajectory data
	Returns:
	DataFrame with columns 'id', 'frame', 'neighbor_id', and 'distance'
	"""
	neighbor_point_col = f"{POINT_COL}_{NEIGHBOR_ID_COL}"
	points = traj_data.data[[ID_COL, FRAME_COL, POINT_COL]]
	neighbor_points = points.rename(
	columns={
	ID_COL: NEIGHBOR_ID_COL,
	POINT_COL: neighbor_point_col,
	}
	)
	matrix = points.merge(
	neighbor_points,
	how="inner",
	on=FRAME_COL,
	)
	matrix = matrix[matrix[ID_COL] != matrix[NEIGHBOR_ID_COL]]
	matrix[DISTANCE_COL] = np.linalg.norm(
	shapely.get_coordinates(matrix[POINT_COL]) - shapely.get_coordinates(matrix[neighbor_point_col]),
	axis=1,
	)
	matrix = matrix.drop(columns=[POINT_COL, neighbor_point_col])
	return matrix.reset_index(drop=True)
	def compute_individual_distances(
	*,
	traj_data: TrajectoryData,
	max_distance: Optional[float] = None,
	) -> pd.DataFrame:
	"""Compute pairwise distances between pedestrians per frame.
	Args:
	traj_data: trajectory data
	max_distance: optional cutoff radius. If provided, only pairs whose
	distance is smaller than or equal to this value are returned.
	Returns:
	DataFrame with columns 'id', 'frame', 'neighbor_id', and 'distance'
	"""
	if max_distance is not None and max_distance < 0:
	raise PedPyValueError("max_distance needs to be greater than or equal to 0.")
	records = []
	points = traj_data.data[[ID_COL, FRAME_COL, POINT_COL]]
	for frame, frame_points in points.groupby(FRAME_COL, sort=False):
	frame_points = frame_points.reset_index(drop=True)
	ids = frame_points[ID_COL].to_numpy()
	geometries = frame_points[POINT_COL].to_list()
	coordinates = shapely.get_coordinates(geometries)
	if len(frame_points) < 2:
	continue
	if max_distance is None:
	for src_idx, dst_idx in itertools.permutations(range(len(frame_points)), 2):
	distance = float(
	np.linalg.norm(coordinates[src_idx] - coordinates[dst_idx])
	)
	records.append(
	{
	ID_COL: ids[src_idx],
	FRAME_COL: frame,
	NEIGHBOR_ID_COL: ids[dst_idx],
	DISTANCE_COL: distance,
	}
	)
	continue
	tree = shapely.STRtree(geometries)
	geometry_to_index = {id(geometry): idx for idx, geometry in enumerate(geometries)}
	for src_idx, geometry in enumerate(geometries):
	candidate_indices = {
	geometry_to_index[id(candidate)]
	for candidate in tree.query(geometry.buffer(max_distance))
	}
	candidate_indices.discard(src_idx)
	for dst_idx in candidate_indices:
	distance = float(
	np.linalg.norm(coordinates[src_idx] - coordinates[dst_idx])
	)
	if distance <= max_distance:
	records.append(
	{
	ID_COL: ids[src_idx],
	FRAME_COL: frame,
	NEIGHBOR_ID_COL: ids[dst_idx],
	DISTANCE_COL: distance,
	}
	)
	return pd.DataFrame.from_records(
	records,
	columns=[ID_COL, FRAME_COL, NEIGHBOR_ID_COL, DISTANCE_COL],
	).reset_index(drop=True)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.52%. Comparing base (7299724) to head (6bf6cda).
⚠️ Report is 3 commits behind head on main.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.