rsasaki0109

rsasaki0109 / gnss_gpu

Public

GPU-accelerated GNSS signal processing library (CUDA + Python)

14
2
100% credibility
Found Apr 06, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

GPU-accelerated GNSS positioning library using particle filters to achieve better urban accuracy than RTKLIB baselines on datasets like UrbanNav and PPC.

How It Works

1
🔍 Discover Better GPS

You find this project while searching for tools to improve GPS accuracy in crowded cities where signals bounce off buildings.

2
📦 Get It Ready

Install it simply so your computer can use its smart location calculations.

3
⚡ Test the Speed

Run quick checks to see how fast it handles location guesses compared to usual methods.

4
🗺️ Feed Your GPS Data

Load your city driving records, and watch colorful particle clouds explore possible positions like a flock finding the best path.

5
📊 See the Results

Compare maps and numbers showing fewer big location mistakes than standard GPS tools.

✅ Reliable City Navigation

Enjoy pinpoint accuracy even in tall buildings, with videos of particles hugging the true path perfectly.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 14 to 14 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is gnss_gpu?

gnss_gpu is a Python library for GPU-accelerated GNSS signal processing, handling everything from raw signal acquisition and tracking to robust positioning in urban canyons. It delivers fast CUDA-backed tools for GNSS SDR workflows, like particle filters that scale to 1M particles for handling multipath and NLOS errors. Users get drop-in Python APIs to process RINEX data, compute positions, and visualize skyplots or vulnerability maps.

Why is it gaining traction?

It crushes CPU baselines—PF variants beat RTKLIB demo5 by 49% RMS on UrbanNav Tokyo sequences, with zero >100m failures and tail robustness up to 1M particles. BVH raytracing delivers 57x speedups on real 3D city models without accuracy loss, making GPU GNSS practical for tough environments. Developers notice the reproducible benchmarks and quick-start pip install turning hours of CPU grind into milliseconds.

Who should use this?

GNSS researchers benchmarking urban positioning algorithms, robotics engineers fusing GPU GNSS with SLAM in cities like Tokyo or Hong Kong, and SDR tinkerers accelerating acquisition/tracking on consumer GPUs. Ideal for teams processing multi-GNSS RINEX from ublox/trimble receivers needing NLOS mitigation without custom C++.

Verdict

Promising alpha library (14 stars, 1.0% credibility) with solid tests (440 passed) and GitHub Pages artifacts—install and repro the UrbanNav wins in minutes. Maturity lags behind RTKLIB, but for GPU-accelerated GNSS processing, it's a sharp tool worth prototyping now.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.