https://github.com/itskalvik/uncertainty-guaranteed-ipp
Informative Path Planning with Guaranteed Estimation Uncertainty (RSS 2026)
https://github.com/itskalvik/uncertainty-guaranteed-ipp
gaussian-processes informative-path-planning robotics sensor-placement sgp-tools
Last synced: 6 days ago
JSON representation
Informative Path Planning with Guaranteed Estimation Uncertainty (RSS 2026)
- Host: GitHub
- URL: https://github.com/itskalvik/uncertainty-guaranteed-ipp
- Owner: itskalvik
- Created: 2025-11-19T18:32:14.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2026-05-27T17:58:18.000Z (16 days ago)
- Last Synced: 2026-05-27T19:27:35.895Z (16 days ago)
- Topics: gaussian-processes, informative-path-planning, robotics, sensor-placement, sgp-tools
- Language: Jupyter Notebook
- Homepage: https://sgp-tools.com
- Size: 136 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Informative Path Planning with Guaranteed Estimation Uncertainty
[](https://arxiv.org/pdf/2602.05198)
[](https://roboticsconference.org/)
Official repository for the paper **"Informative Path Planning with Guaranteed Estimation Uncertainty"**, published at **Robotics: Science and Systems (RSS), 2026**.
This repository contains the benchmarking scripts used to produce the figures and experimental results presented in the paper. Our core IPP planners—**GreedyCover** and **GCBCover**—are integrated into the [SGP-Tools](https://www.sgp-tools.com/) library.
---
## 🛠 Repository Structure
```text
.
├── datasets/ # SRTM subsets (N02E021, N17E073, N45W123, N47W124)
├── benchmark.py # Main benchmarking script, also generates IPP solution figures
├── plot.py # Generates benchmark plots from json files generated by benchmark.py
├── cover.ipynb # Reproduces the paper's cover page IPP solutions
├── fov.ipynb # Coverage-map visualizations for the methods section
├── requirements.txt # Project dependencies
└── README.md
```
---
## 🚀 Getting Started
### 1. Environment Setup
We recommend using a virtual environment (`venv`) or `conda` to manage dependencies.
```bash
# Clone the repository
git clone https://github.com/itskalvik/uncertainty-guaranteed-ipp.git
cd uncertainty-guaranteed-ipp
# Install dependencies
python -m pip install -r requirements.txt
```
> 📝 Note
> This installs `sgptools`, which relies on **TensorFlow** and **GPflow**. A GPU is not required.
### 2. LaTeX Dependency
The plotting scripts (`benchmark.py` and `plot.py`) use LaTeX for publication-quality rendering.
* **If you have LaTeX installed:** No changes needed.
* **If you do NOT have LaTeX:** Edit the scripts to set `matplotlib.rcParams["text.usetex"] = False`.
---
## 📊 Running Benchmarks
### Step 1: Run the IPP Planners
Execute the benchmark on one of the provided datasets by sweeping through target variance thresholds (expressed as ratios of the initial pilot model's max variance).
```bash
python3 benchmark.py ./datasets/N47W124.npy --variance-ratios 0.9 0.8 0.7 0.6 0.5
```
**Methods included:** `HexCover`, `GreedyCover`, `GCBCover`, and `GCBCover-Dist`.
**Outputs:** Solution figures and a comprehensive `results.json` file.
### Step 2: Generate Plots
Once the benchmark completes, visualize the performance metrics (MSE, SMSE, Runtime, Distance, etc.):
```bash
# Set the specific results file generated in Step 1
python3 plot.py N47W124/Attentive/results.json
```
---
## 📝 Key Algorithms
The following planners are the primary contributions of this work:
* **GreedyCover:** An efficient greedy algorithm for near-optimal IPP with uncertainty guarantees.
* **GCBCover:** Balances information gain and travel costs to solve IPP with uncertainty guarantees under routing constraints.
For standalone use of these planners in your own projects, please refer to the [SGP-Tools documentation](https://www.sgp-tools.com/tutorials/uncertainty_guaranteed_IPP.html).
---
## 🎓 Citation
If you find this work useful for your research, please cite our RSS 2026 paper:
```bibtex
@inproceedings{JakkalaAOA26,
author = {Kalvik Jakkala and Saurav Agarwal and Jason O'Kane and Srinivas Akella},
title = {Informative Path Planning with Guaranteed Estimation Uncertainty},
booktitle = {Robotics: Science and Systems (RSS)},
year = {2026},
url = {https://www.itskalvik.com/publication/uncertainty-guaranteed-ipp/}
}
```
---
## ⚖️ Notes on Reproducibility
* **Seeds:** `benchmark.py` sets `numpy` and `tensorflow` random seeds.
* **Hardware:** Minor variations in floating-point math may occur if running on a GPU; however, the algorithmic trends remain consistent.