libgnss++ — Modern C++ GNSS/RTK/PPP/CLAS Toolkit

Native non-GUI GNSS stack in modern C++17 with built-in SPP, RTK, PPP, CLAS/MADOCA, RTCM, UBX, and direct QZSS L6 handling.

The point of this repo is simple: ship a usable GNSS toolchain without depending on an external RTKLIB runtime.

If this repo is useful, star it.

Contribution and PR workflow: CONTRIBUTING.md Architecture notes: docs/architecture.md Documentation index: docs/index.md

CLAS Performance vs CLASLIB

QZSS CLAS (Centimeter-Level Augmentation Service) PPP from raw L6 binary, 2019-08-27 static dataset (TRM59800.80 antenna), 1 hour (3599 epochs):

Metric	gnssplusplus `--claslib-parity`	CLASLIB
Matched fixed epochs	3594 / 3599 (99.86%)	3594 / 3599 (99.86%)
RMS 3D (fixed-only)	3.57 mm	7.29 mm
3D bias (mean offset)	1.66 mm	4.84 mm
RMS East	1.15 mm	1.52 mm
RMS North	1.21 mm	0.92 mm
RMS Up	3.15 mm	7.07 mm
Mean E / N / U	-0.72 / +0.93 / +1.17 mm	+0.65 / -0.59 / +4.76 mm
First fix epoch	epoch 6	epoch 6
CLASLIB runtime link	not required	required
Parity depth	17 helpers at 1e-6 m vs CLASLIB oracle	reference

CLASLIB 2D	gnssplusplus 2D

gnssplusplus achieves 51% lower RMS 3D and ~3x tighter 3D bias than upstream CLASLIB on the same 1-hour window while keeping the same fix rate, with no CLASLIB runtime dependency on the default path. The ClasnatParity GoogleTest suite pins 17 core helpers (windupcorr, antmodel, ionmapf, prectrop, corrmeas, satpos_ssr, tidedisp, eph2clk, eph2pos, geodist, satantoff, compensatedisp, trop_grid_data, filter, lambda, tropmodel, stec_grid_data) to 1e-6 m parity against the CLASLIB C source.

Opt-ins:

-DCLASLIB_PARITY_LINK=ON + --claslib-bridge: delegate to upstream CLASLIB postpos() linked as a static library (oracle mode)
--legacy-strict-parity: iter13-era non-native strict OSR path (regression reference)

See docs/clas_port_architecture.md for the port design and docs/clas_validated_datasets.md for the validated dataset set.

RTK Performance vs RTKLIB (demo5)

gnssplusplus develop (post PR #19–#23) dominates RTKLIB demo5 on the PPC-Dataset Tokyo and Nagoya urban runs across Fix count, Fix rate, and precision — with no Phase 2 opt-in flags. UrbanNav Tokyo Odaiba is dominated on Fix count, Hp95, and Vp95; Hmed sits within 9 cm of demo5 once --enable-wide-lane-ar --wide-lane-threshold 0.10 is opted in.

All runs below use --mode kinematic --preset low-cost --match-tolerance-s 0.25.

PPC Tokyo (kinematic, low-cost preset, no Phase 2 flags)

Run	gnssplusplus Fix / rate	RTKLIB Fix / rate	Hmed (m)	Vp95 (m)
run1	3572 / 81.26%	2418 / 30.52%	0.037 vs 1.567 (42×)	1.259 vs 36.703 (29×)
run2	4674 / 80.12%	2127 / 27.58%	0.016 vs 0.835 (52×)	0.313 vs 42.624 (136×)
run3	7516 / 86.84%	5778 / 40.55%	0.012 vs 0.666 (56×)	0.137 vs 24.521 (179×)

PPC Nagoya (same preset)

Run	Fix delta	rate delta	Hmed delta
run1	+1743	+58.03 pp	9× better
run2	+1735	+64.00 pp	10× better
run3	+154	+50.16 pp	44× better

UrbanNav Tokyo Odaiba (kinematic, low-cost preset)

Config	Fix	Rate	Hmed (m)	Hp95 (m)	Vp95 (m)
RTKLIB demo5	595	7.22%	0.707	27.878	45.212
gnssplusplus default	1268 (+673)	36.98%	1.707	19.585	25.495
gnssplusplus `--enable-wide-lane-ar --wide-lane-threshold 0.10`	818 (+223)	33.65%	0.799 (9 cm gap)	19.971	26.429

PPC Tokyo + Nagoya need no Phase 2 flags. On Odaiba, --enable-wide-lane-ar --wide-lane-threshold 0.10 is the precision optimum.

RTKLIB 2D	libgnss++ 2D

Phase 2 opt-in tuning gates

The default RTK pipeline already dominates demo5 on the production datasets above. Five additional gates ship default-off for situations where you want to push further on precision-vs-fix-count tradeoffs. All are byte-identical to the default behavior unless explicitly enabled.

Flag	Purpose	Default
`--ar-policy {extended\|demo5-continuous}`	AR extras gate. `demo5-continuous` disables relaxed-hold-ratio / subset-fallback / hold-fix / Q-regularization for demo5-style continuous ambiguity tracking.	`extended`
`--max-hold-div <m>`	Reject fix if the hold-state diverges from float by more than N meters.	`0` (disabled)
`--max-pos-jump <m>`	Reject fix if the epoch-to-epoch position jump exceeds N meters.	`0` (disabled)
`--max-consec-float-reset <N>`	Auto-reset ambiguities after N consecutive float epochs.	`0` (disabled)
`--max-postfix-rms <m>`	Reject fix if the L1 post-fix DD phase residual RMS exceeds N meters.	`0` (disabled)
`--enable-wide-lane-ar` + `--wide-lane-threshold <cycle>`	Pre-compute MW wide-lane integers and inject them as Kalman constraints into the LAMBDA search. Halves Hmed on Odaiba at the cost of ~35% Fix count.	`false` / `0.25`

These were added in PR #19–#23. On PPC Tokyo and Nagoya the defaults already win, so leave them off. On Odaiba (or other urban multipath sets), --enable-wide-lane-ar --wide-lane-threshold 0.10 is the precision optimum.

Docs

Local docs site:

python3 -m pip install -r requirements-docs.txt
python3 -m mkdocs serve

Docker

Build the runtime image:

docker build -t libgnsspp:latest .

Pull the published image:

docker pull ghcr.io/rsasaki0109/gnssplusplus-library:develop

Run the CLI against a mounted workspace or dataset directory:

docker run --rm -it \
  -v "$PWD:/workspace" \
  libgnsspp:latest \
  solve --rover /workspace/data/rover_kinematic.obs \
  --base /workspace/data/base_kinematic.obs \
  --nav /workspace/data/navigation_kinematic.nav \
  --out /workspace/output/docker_rtk.pos

Serve the local web UI from inside the container:

docker run --rm -it \
  -p 8085:8085 \
  -v "$PWD:/workspace" \
  libgnsspp:latest \
  web --host 0.0.0.0 --port 8085 --root /workspace

The image installs the gnss dispatcher, Python helpers, and libgnsspp Python package, but it does not embed the repo sample datasets. Mount your source tree or your own dataset directory.

Run the web UI with Compose:

docker compose up gnss-web

Override the image if you want a local or tagged build:

LIBGNSSPP_IMAGE=ghcr.io/rsasaki0109/gnssplusplus-library:v0.1.0 docker compose up gnss-web

What You Get

Native solvers: SPP, RTK, PPP, CLAS-style PPP
Native protocols: RINEX, RTCM, UBX, direct QZSS L6
Raw/log tooling: NMEA, NovAtel, SBP, SBF, Trimble, SkyTraq, BINEX
Product tooling: fetch-products, ionex-info, dcb-info
Analysis tooling: visibility, visibility-plot, and moving-base-plot for az/el/SNR exports plus moving-base/visibility PNG quick-looks
Moving-base tooling: moving-base-prepare plus moving-base-signoff for real bag/replay/live validation
One CLI entrypoint: gnss spp, solve, ppp, visibility, stream, convert, live, rcv
Local web UI: gnss web for benchmark snapshots, live/moving-base/PPP-product sign-offs, 2D trajectories, visibility views, artifact bundles, receiver status, and artifact links
Built-in sign-off scripts and checked-in benchmark artifacts
CMake install/export, Python bindings, and ROS2 playback node

Quick Start

Build

cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build -j

First solutions

python3 apps/gnss.py spp \
  --obs data/rover_static.obs \
  --nav data/navigation_static.nav \
  --out output/spp_solution.pos

python3 apps/gnss.py solve \
  --rover data/short_baseline/TSK200JPN_R_20240010000_01D_30S_MO.rnx \
  --base data/short_baseline/TSKB00JPN_R_20240010000_01D_30S_MO.rnx \
  --nav data/short_baseline/BRDC00IGS_R_20240010000_01D_MN.rnx \
  --mode static \
  --out output/rtk_solution.pos

python3 apps/gnss.py ppp \
  --static \
  --obs data/rover_static.obs \
  --nav data/navigation_static.nav \
  --out output/ppp_solution.pos

python3 apps/gnss.py visibility \
  --obs data/rover_static.obs \
  --nav data/navigation_static.nav \
  --csv output/visibility.csv \
  --summary-json output/visibility_summary.json \
  --max-epochs 60

python3 apps/gnss.py replay \
  --rover-rinex data/rover_kinematic.obs \
  --base-rinex data/base_kinematic.obs \
  --nav-rinex data/navigation_kinematic.nav \
  --mode moving-base \
  --out output/moving_base_replay.pos \
  --max-epochs 20

Inspect receiver logs

python3 apps/gnss.py ubx-info \
  --input logs/session.ubx \
  --decode-observations

python3 apps/gnss.py sbf-info \
  --input logs/session.sbf \
  --decode-pvt \
  --decode-lband \
  --decode-p2pp

Useful commands

Command	Purpose
`gnss spp`	Batch SPP from rover/nav RINEX
`gnss solve`	Batch RTK from rover/base/nav RINEX
`gnss ppp`	Batch PPP from rover RINEX plus nav or precise products
`gnss visibility`	Export azimuth/elevation/SNR visibility rows and summary JSON from rover/nav RINEX
`gnss visibility-plot`	Render a visibility CSV into a polar/elevation PNG quick-look
`gnss moving-base-plot`	Render a moving-base solution/reference pair into a baseline/heading PNG quick-look
`gnss fetch-products`	Fetch and cache `SP3`/`CLK`/`IONEX`/`DCB` files from local or remote sources
`gnss moving-base-prepare`	Extract rover/base UBX plus reference CSV from a ROS2 moving-base bag
`gnss scorpion-moving-base-signoff`	Prepare and validate the public SCORPION moving-base ROS2 bag through replay
`gnss stream`	Inspect and relay RTCM over file, NTRIP, TCP, or serial
`gnss convert`	Convert RTCM or UBX into simple RINEX outputs
`gnss ubx-info`	Inspect `NAV-PVT`, `RAWX`, `SFRBX` from file or serial
`gnss sbf-info`	Inspect Septentrio SBF `PVTGeodetic`, `LBandTrackerStatus`, `P2PPStatus` from file or serial
`gnss novatel-info`	Inspect NovAtel ASCII/Binary `BESTPOS` and `BESTVEL` logs
`gnss nmea-info`	Inspect `GGA` and `RMC` NMEA logs from file or serial
`gnss ionex-info`	Inspect `IONEX` header, map count, grid metadata, and auxiliary DCB blocks
`gnss dcb-info`	Inspect `Bias-SINEX` or auxiliary DCB product contents
`gnss qzss-l6-info`	Inspect direct QZSS L6 frames and export Compact SSR payloads
`gnss social-card`	Regenerate the Odaiba share image
`gnss short-baseline-signoff`	Static RTK sign-off
`gnss rtk-kinematic-signoff`	Kinematic RTK sign-off
`gnss ppp-static-signoff`	Static PPP sign-off
`gnss ppp-kinematic-signoff`	Kinematic PPP sign-off
`gnss ppp-products-signoff`	Static, kinematic, or PPC PPP sign-off with fetched SP3/CLK/IONEX/DCB products, optional MALIB delta gates, and comparison CSV/PNG artifacts
`gnss live-signoff`	Realtime/error-handling sign-off for recorded RTCM/UBX live inputs
`gnss ppc-demo`	External PPC-Dataset RTK/PPP verification against `reference.csv`
`gnss ppc-rtk-signoff`	Fixed RTK sign-off profiles for PPC Tokyo/Nagoya, with optional RTKLIB side-by-side gates
`gnss moving-base-signoff`	Real moving-base replay/live sign-off against per-epoch base/rover reference coordinates
`gnss odaiba-benchmark`	End-to-end Odaiba benchmark pipeline
`gnss web`	Local browser UI for summary JSON, live/moving-base/PPP-product sign-offs, `.pos` trajectories, moving-base/visibility plots and histories, receiver status, and artifact/provenance links

See all commands:

python3 apps/gnss.py --help

Local web UI

python3 apps/gnss.py web \
  --port 8085 \
  --rcv-status output/receiver.status.json

Then open http://127.0.0.1:8085 to inspect Odaiba metrics, live/moving-base/PPP-product sign-offs, 2D trajectories, moving-base and visibility plots, moving-base history, PPC summaries, receiver status, and linked artifact bundles in a browser. The PPP products table links directly to fetched products, MALIB .pos, comparison CSV/PNG artifacts, and dataset provenance.

Long-running dashboard commands can also read TOML config files. See configs/web.example.toml, configs/live_signoff.example.toml, configs/moving_base_signoff.example.toml, configs/ppc_rtk_signoff.example.toml, and configs/ppp_products_ppc.example.toml, then pass --config-toml <file>.

Container form:

docker run --rm -it -p 8085:8085 -v "$PWD:/workspace" \
  libgnsspp:latest web --host 0.0.0.0 --port 8085 --root /workspace

Real moving-base sign-off

gnss solve, gnss replay, and gnss live accept --mode moving-base. For real moving-base datasets, use gnss moving-base-signoff with a reference CSV carrying per-epoch base/rover ECEF coordinates. The repo does not ship a bundled moving-base dataset, so this command is intended for external real logs.

python3 apps/gnss.py moving-base-prepare \
  --input /datasets/moving_base/2023-06-14T174658Z.zip \
  --rover-ubx-out output/moving_base_rover.ubx \
  --base-ubx-out output/moving_base_base.ubx \
  --reference-csv output/moving_base_reference.csv \
  --summary-json output/moving_base_prepare.json

python3 apps/gnss.py fetch-products \
  --date 2023-06-14 \
  --preset brdc-nav \
  --summary-json output/moving_base_products.json

python3 apps/gnss.py moving-base-signoff \
  --solver replay \
  --rover-ubx output/moving_base_rover.ubx \
  --base-ubx output/moving_base_base.ubx \
  --nav-rinex ~/.cache/libgnsspp/products/nav/2023/165/BRDC00IGS_R_20231650000_01D_MN.rnx \
  --reference-csv output/moving_base_reference.csv \
  --summary-json output/moving_base_summary.json \
  --require-fix-rate-min 90 \
  --require-p95-baseline-error-max 1.0 \
  --require-realtime-factor-min 1.0 \
  --max-epochs 120

python3 apps/gnss.py scorpion-moving-base-signoff \
  --input-url https://zenodo.org/api/records/8083431/files/2023-06-14T174658Z.zip/content \
  --summary-json output/scorpion_moving_base_summary.json \
  --require-matched-epochs-min 100 \
  --require-fix-rate-min 80

python3 apps/gnss.py moving-base-signoff \
  --config-toml configs/moving_base_signoff.example.toml

python3 apps/gnss.py live-signoff \
  --config-toml configs/live_signoff.example.toml

Product-driven PPP

python3 apps/gnss.py fetch-products \
  --date 2024-01-02 \
  --preset igs-final \
  --preset ionex \
  --preset dcb \
  --summary-json output/products.json

python3 apps/gnss.py ppp-static-signoff \
  --fetch-products \
  --product-date 2024-01-02 \
  --product sp3=https://cddis.nasa.gov/archive/gnss/products/{gps_week}/COD0OPSFIN_{yyyy}{doy}0000_01D_05M_ORB.SP3.gz \
  --product clk=https://cddis.nasa.gov/archive/gnss/products/{gps_week}/COD0OPSFIN_{yyyy}{doy}0000_01D_30S_CLK.CLK.gz \
  --product ionex=https://cddis.nasa.gov/archive/gnss/products/ionex/{yyyy}/{doy}/COD0OPSFIN_{yyyy}{doy}0000_01D_01H_GIM.INX.gz \
  --product dcb=https://cddis.nasa.gov/archive/gnss/products/bias/{yyyy}/CAS0MGXRAP_{yyyy}{doy}0000_01D_01D_DCB.BSX.gz \
  --summary-json output/ppp_static_summary.json

python3 apps/gnss.py ppp-kinematic-signoff \
  --max-epochs 120 \
  --require-common-epoch-pairs-min 120 \
  --require-reference-fix-rate-min 95 \
  --require-converged \
  --require-convergence-time-max 300 \
  --require-mean-error-max 7 \
  --require-p95-error-max 7 \
  --require-max-error-max 7 \
  --require-mean-sats-min 18 \
  --require-ppp-solution-rate-min 100

python3 apps/gnss.py ppp-products-signoff \
  --config-toml configs/ppp_products_ppc.example.toml

python3 apps/gnss.py ppc-rtk-signoff \
  --config-toml configs/ppc_rtk_signoff.example.toml

Benchmark Snapshot

UrbanNav Tokyo Odaiba

Dataset: UrbanNav Tokyo Odaiba (2018-12-19, Trimble rover/base, ~170 m baseline). Comparison baseline: RTKLIB.

Current checked-in snapshot (kinematic, low-cost preset):

RTKLIB demo5: Fix 595 / Rate 7.22% / Hmed 0.707 m / Hp95 27.878 m / Vp95 45.212 m
libgnss++ default: Fix 1268 (+673) / Rate 36.98% / Hmed 1.707 m / Hp95 19.585 m / Vp95 25.495 m
libgnss++ --enable-wide-lane-ar --wide-lane-threshold 0.10: Fix 818 (+223) / Rate 33.65% / Hmed 0.799 m (9 cm gap) / Hp95 19.971 m / Vp95 26.429 m

libgnss++ dominates Fix count, Hp95, and Vp95 at any config; Hmed is the sole demo5 edge under default flags, and Phase 2 wide-lane AR closes it to 9 cm.

RTKLIB 2D	libgnss++ 2D

More artifacts:

Full comparison figure
Scorecard
Summary JSON
Optional side-by-side PPP benchmark path: JAXA-SNU/MALIB
Additional low-cost GNSS RTK/PPP reference implementation: rtklibexplorer/RTKLIB

Other checked sign-offs

Mixed-GNSS short-baseline RTK
Mixed-GNSS kinematic RTK
Static PPP
Kinematic PPP
CLAS-style PPP from compact sampled SSR and raw QZSS L6

External dataset demo

PPC-Dataset can be verified directly from an extracted dataset tree:

python3 apps/gnss.py ppc-demo \
  --dataset-root /datasets/PPC-Dataset \
  --city tokyo \
  --run run1 \
  --solver rtk \
  --require-realtime-factor-min 1.0 \
  --summary-json output/ppc_tokyo_run1_rtk_summary.json

python3 apps/gnss.py ppc-rtk-signoff \
  --dataset-root /datasets/PPC-Dataset \
  --city tokyo \
  --rtklib-bin /path/to/rnx2rtkp \
  --summary-json output/ppc_tokyo_run1_rtk_signoff.json

Dataset source: taroz/PPC-Dataset

Install And Package

cmake --install build --prefix /opt/libgnsspp

Installed layout includes:

bin/gnss
native binaries such as gnss_spp, gnss_solve, gnss_ppp, gnss_stream
Python command wrappers and sign-off scripts
scripts/ asset generators
lib/cmake/libgnsspp/libgnssppConfig.cmake
lib/pkgconfig/libgnsspp.pc
Python package libgnsspp

Examples:

# pkg-config
pkg-config --cflags --libs libgnsspp

# source the installed dispatcher
/opt/libgnsspp/bin/gnss social-card \
  --lib-pos output/rtk_solution.pos \
  --rtklib-pos output/driving_rtklib_rtk.pos \
  --reference-csv data/driving/Tokyo_Data/Odaiba/reference.csv \
  --output docs/driving_odaiba_social_card.png

Python And ROS2

Python bindings expose:

RINEX header and epoch inspection
.pos loading and solution statistics
coordinate conversion helpers
file-based SPP, PPP, and RTK solve helpers

ROS2 support includes a playback node that publishes .pos files as:

sensor_msgs/NavSatFix
geometry_msgs/PoseStamped
nav_msgs/Path
solution status and satellite-count telemetry

Tests And Dogfooding

Run the full non-GUI regression set:

ctest --test-dir build --output-on-failure

Important checks already covered in-tree:

solver/unit tests
live realtime/error-handling regression
benchmark/image generation tests
installed-prefix packaging smoke tests
installed gnss social-card dogfooding
installed feature-overview image generation
Python bindings smoke tests
ROS2 node smoke tests

Data

Bundled samples live under:

data/
data/short_baseline/
data/driving/Tokyo_Data/Odaiba/

Generated benchmark outputs live under:

output/
docs/

Scope

This repo is intentionally focused on a strong non-GUI GNSS stack.

It already covers:

native RTK, PPP, CLAS, RTCM, UBX, QZSS L6
installed CLI tooling
benchmarks, sign-off scripts, and README asset generation

It is still not marketed as a perfect RTKLIB drop-in replacement. The remaining gaps are about scope breadth, not the core non-GUI workflow shipped here.

License

MIT License. See LICENSE.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

libgnss++ — Modern C++ GNSS/RTK/PPP/CLAS Toolkit

CLAS Performance vs CLASLIB

RTK Performance vs RTKLIB (demo5)

PPC Tokyo (kinematic, low-cost preset, no Phase 2 flags)

PPC Nagoya (same preset)

UrbanNav Tokyo Odaiba (kinematic, low-cost preset)

Phase 2 opt-in tuning gates

Docs

Docker

What You Get

Quick Start

Build

First solutions

Inspect receiver logs

Useful commands

Local web UI

Real moving-base sign-off

Product-driven PPP

Benchmark Snapshot

UrbanNav Tokyo Odaiba

Other checked sign-offs

External dataset demo

Install And Package

Python And ROS2

Tests And Dogfooding

Data

Scope

License

FilesExpand file tree

README.md

Latest commit

History

README.md

File metadata and controls

libgnss++ — Modern C++ GNSS/RTK/PPP/CLAS Toolkit

CLAS Performance vs CLASLIB

RTK Performance vs RTKLIB (demo5)

PPC Tokyo (kinematic, low-cost preset, no Phase 2 flags)

PPC Nagoya (same preset)

UrbanNav Tokyo Odaiba (kinematic, low-cost preset)

Phase 2 opt-in tuning gates

Docs

Docker

What You Get

Quick Start

Build

First solutions

Inspect receiver logs

Useful commands

Local web UI

Real moving-base sign-off

Product-driven PPP

Benchmark Snapshot

UrbanNav Tokyo Odaiba

Other checked sign-offs

External dataset demo

Install And Package

Python And ROS2

Tests And Dogfooding

Data

Scope

License