Official Dexmate manual — For this setup, you generally only need the Network and Software Setup sections.
Chinese manual · English manual
Vega-1U setup kit: robot-exported gripper library, dual-wrist camera initialization/API, pose scripts, pinned camera/CAN drivers, and an illustrated bilingual GitHub Pages manual.
Target: Orin Nano 8GB (p3767-0005 / p3768), L4T 36.5.0, Linux 5.15.185-tegra, aarch64, system Python 3.10, firmware 0.5.x. Prepare the robot's own certificate and matching vendor baseline first: ZED SDK 5.2.3, ZED Link Duo 1.4.3-LI-MAX96712-L4T36.5.0 and dexsensor 0.7.6. Exact vendor filenames/hashes: vendor-installers.json.
git clone https://github.com/intelligent-control-lab/dexmate-setup.git
cd dexmate-setup
# Use THIS robot's serial, not the source robot's identity.
sudo /usr/bin/python3 scripts/install.py --apply \
--robot-name dm/YOUR_SERIAL-1u --user dexmate \
--install-deps --camera-boot --activate-camera-boot --configure-headReview the generated boot entry, then reboot manually with a display/keyboard available. Log in again for group membership, then:
source /opt/dexmate-setup/env.sh
cd ~/dexmate-setup # or your actual clone path
python3 scripts/doctor.pyThe installer preserves the target robot's root disk arguments and original boot entries,
backs up replaced files, and does not start services, reboot, or command motors.
Omit --activate-camera-boot to stage the camera boot entry without selecting it.
The dependency installation needs Internet access; this is not an offline factory image.
The installer enables wrist-camera-init.service for boot-time initialization. Wrist capture
starts on demand through WristCameras(). The head uses the vendor's dexsensor.service;
--configure-head configures it but does not enable it. After installation, enable both:
sudo systemctl daemon-reload
sudo systemctl enable dexsensor.service wrist-camera-init.serviceSee the manual's camera startup and troubleshooting instructions.
| Path | Purpose |
|---|---|
software/gripper/ |
Grippers/Motor library and connection/self-test tool |
software/wrist_cameras/ |
On-demand dual RGB API, initialization and trigger worker |
software/poses/ |
Latest robot-exported ready/rest, head and pose scripts |
drivers/ |
Exact-kernel modules, overlay, checksums and rebuild inputs |
config/ |
CAN/wrist services, tmpfiles and captured dexsensor template |
scripts/ |
Installer, initrd builder, read-only doctor, site generator |
docs/ |
Original manual, hardware setup, calibration/URDF reference, Chinese translation, photos and videos |
provenance/ |
Original source hashes, package versions and validation limits |
Hardware and control measurements · TCP and camera extrinsics
- Supplied package models: vega_1u.urdf and vega_1u_gripper.urdf.
- Clean calibration pair, pivot-fit result, and earlier Y-adjusted pair.
- Reference matrices and file hashes, including the separate custom URDF nominal TCP.
The supplied original is preserved in manual/vega1umanual.html. The website keeps its
layout, commands and examples. Part B contains mandatory chapters B1–B8: gripper hardware,
PCB/camera hardware, network, certificates, workstation communication, libraries, verification,
and cameras. Part C contains advanced references, hardware/control information and TCP/extrinsics.
Chinese prose for the original chapters is in manual/zh.json.
Edit manual/hardware.en.html / manual/hardware.zh.html for hardware setup and
manual/reference.en.html / manual/reference.zh.html for calibration references.
Camera startup and troubleshooting additions are in manual/camera-help.en.html /
manual/camera-help.zh.html. Chapter order and numbering are set in scripts/build_docs.py. Then:
python3 scripts/build_docs.py
python3 -m unittest discover -s tests -v
python3 scripts/check_site.py
python3 scripts/check_reference.py
python3 -m http.server 8000 --directory docsGitHub Pages publishes main /docs. No frontend build system or API keys are required.