https://github.com/winstxnhdw/scipath
A path interpolation wrapper for SciPy's CubicSpline.
https://github.com/winstxnhdw/scipath
cubic-splines interpolation path-planning scipy uv
Last synced: about 1 month ago
JSON representation
A path interpolation wrapper for SciPy's CubicSpline.
- Host: GitHub
- URL: https://github.com/winstxnhdw/scipath
- Owner: winstxnhdw
- Created: 2021-04-07T21:49:38.000Z (about 4 years ago)
- Default Branch: main
- Last Pushed: 2025-03-17T21:15:17.000Z (about 1 month ago)
- Last Synced: 2025-03-17T21:42:04.580Z (about 1 month ago)
- Topics: cubic-splines, interpolation, path-planning, scipy, uv
- Language: Python
- Homepage:
- Size: 2.13 MB
- Stars: 0
- Watchers: 2
- Forks: 1
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# scipath
[](https://www.python.org/)
[](https://github.com/winstxnhdw/scipath/actions/workflows/main.yml)
[](https://github.com/winstxnhdw/scipath/actions/workflows/formatter.yml)
> [!WARNING]\
> You should never use `scipath` in production. The API may change at any time and there are no plans for releases with semantic versioning. This package is intended for generating animations in my legacy autonomous vehicle projects.
`scipath` is a lightweight typesafe wrapper for SciPy's [CubicSpline](https://docs.scipy.org/doc/scipy/reference/generated/scipy.interpolate.CubicSpline.html). This wrapper simplifies the interpolation of coarse path data and allows the user to solely compute the profile of the path, such as path curvature and yaw. For large paths, it can be up to 300x faster than Atsushi Sakai's [PyCubicSpline](https://github.com/AtsushiSakai/pycubicspline).
## Installation
```bash
pip install git+https://github.com/winstxnhdw/scipath
```
## Usage
`scipath` is fast because it heavily relies on NumPy broadcasting and lazily computes the expensive $p'(x)$ and $p''(x)$ only once. The following example demonstrates how to use `scipath`.
```python
from scipath import Profile, create_cubic_path_2d
points = [(0, 0), (1, 1), (2, 0), (3, 1)]
path, yaw, curvature = create_cubic_path_2d(points, profile=Profile.ALL)
```
To reduce the necessary computation required, you can specify a subset of the profiles you want to compute.
```python
from scipath import Profile, create_cubic_path_2d
points = [(0, 0), (1, 1), (2, 0), (3, 1)]
create_cubic_path_2d(points, profile=Profile.PATH)
create_cubic_path_2d(points, profile=Profile.YAW)
create_cubic_path_2d(points, profile=Profile.CURVATURE)
create_cubic_path_2d(points, profile=Profile.NO_CURVATURE)
create_cubic_path_2d(points, profile=Profile.NO_YAW)
create_cubic_path_2d(points, profile=Profile.NO_PATH)
```
## Benchmarks
In our benchmarks, we compare against Atsushi Sakai's [PyCubicSpline](https://github.com/AtsushiSakai/pycubicspline), and found that `scipath` is ~250x faster.
```mermaid
---
config:
xyChart:
height: 200
titleFontSize: 15
themeVariables:
xyChart:
backgroundColor: "#00000000"
---
xychart-beta horizontal
x-axis [scipath, pycspline]
y-axis "Compute Time (ms)" 0 --> 260
bar [1.893548, 251.382491]
```
You may replicate the benchmark results with the following.
```bash
git clone --recursive https://github.com/winstxnhdw/scipath.git
cd scipath
uv run benchmarks/benchmark.py
```