QGC_KevinJiang v2.1.0

QGC_KevinJiang v2.1.0

Release date: 2026-10-24

QGC_KevinJiang v2.1.0 is a Windows release for GRobot project members, based on QGroundControl v5.1.5. It includes all changes from the unreleased 2.0.1 candidate. Full release notes (Chinese) are attached as v2.1.0.zh-CN.md.

Highlights

• Arming and physical outputs: thruster direct output tests and arm requests are mutually exclusive; a test cannot start, restore, or save while an arm/disarm or motor-test command sent through QGC is pending or the arm state is unconfirmed. Returning PWM to neutral and restoring the original function are always available. The ArduSub motor page works again.
• Calibration: when the flight controller stops responding, a local wait can be ended after a timeout; every calibration can be cancelled.
• Joystick: changing extension or auxiliary axis settings clears joystick calibration and releases RC override.
• Video and AI: RTSP reconnects even when it fails before the first frame; changing AI port, camera URLs, or enabled channels stops detection first.
• Installer: during in-place upgrade the previous uninstaller runs from a temporary copy and only in its own directory, so it no longer deletes files of the new installation or leaves the uninstaller scheduled for deletion.
• From 2.0.1: neutral confirmation from the outputs reported by the flight controller, joystick thread-safety fixes, camera credentials hidden in status panels, and attitude and depth shown only after the vehicle reports them.

Upgrading from 2.0.x

1. Quit QGC, back up the whole %APPDATA%\KevinJiang\ directory, and keep the previous installer.
2. Run QGC_v2.1.0_KevinJiang-installer.exe into the existing installation directory; uninstalling first is not needed.
3. Confirm version 2.1.0 at the bottom of the menu or under Application Settings → Help, and check the video URLs, application settings, and joystick mapping.

If you upgraded before and have not restarted since, restart Windows before installing 2.1.0.

Verified artifact

• Release audit 20261024-101221-031 succeeded with all eight steps passing; installer version 2.1.0, product version 2.1.0.0.
• Size: 137859735 bytes
• SHA-256: 729508CDF8CC2B26E198EDB29F28661E3B978B6B6379336632497D6BC1FECDAF
• Authenticode status: not signed

Important boundary

• The arming, calibration, and joystick changes are covered by automated tests only and have not been verified on a real vehicle. Before the first thruster or servo test, make sure the vehicle is in a safe position with no people or obstacles within its range of motion.
• The in-place upgrade was verified on the maintainer's machine only.
• Python, CUDA, and AI model files are not bundled; configure them on each machine.

Source

Published on GitHub on 2026-10-27; it was previously distributed to GRobot members only. Tag qgc-kevinjiang-v2.1.0 points at 564b7ab6c, the last commit before the successful packaging run.

This is not an official QGroundControl release. The project retains upstream license and provenance information.

QGC_KevinJiang v2.0.0

QGC_KevinJiang v2.0.0

Release date: 2026-10-07

QGC_KevinJiang v2.0.0 is a Windows release for GRobot project members, based on QGroundControl v5.1.5. Full release notes (Chinese) are attached as v2.0.0.zh-CN.md.

Highlights

• Upgraded the upstream base to QGroundControl v5.1.5, together with Qt and GStreamer, keeping the DeepShark four-camera video, AI detection, and ArduSub customizations.
• The main toolbar shows arm/disarm and heartbeat checks directly; joystick setup is reorganized into four pages (general, button assignment, calibration, advanced).
• Completed Chinese translations of commonly used screens, added an exit confirmation, and merged duplicate connection-alert settings entries.
• Improved video decoding edge cases, texture isolation between streams, and frame backlog and output switching while the main window is resized.

Upgrading from 1.5.0

1. Quit QGC, back up the whole %APPDATA%\KevinJiang\ directory, and keep the 1.5.0 installer.
2. Run QGC_v2.0.0_KevinJiang-installer.exe into the existing installation directory; uninstalling first is not needed.
3. Confirm the version at the bottom of the menu or under Application Settings → Help, and check the video URLs, application settings, and joystick mapping.

To roll back, run the kept 1.5.0 installer into the same directory and restore the configuration backup before first starting 1.5.0. A plain uninstall may delete the configuration and is not recommended for rollback.

Verified artifact

• Release audit 20261007-192551-204 succeeded with all eight steps passing; installer version 2.0.0, product version 2.0.0.0.
• Size: 137806770 bytes
• SHA-256: 9B324EA1C6DE9DF97CF397DF04DD348F683CBB2ACBF3CF96EDA141242A87DED1
• Authenticode status: not signed

Known limitations

• Real joystick operation, cameras, flight controller, and long-running field use still need testing by members; the release does not claim to have resolved every root cause of the field hang during window resizing.
• Python, CUDA, and AI model files are not bundled; configure them on each machine.

Source

Published on GitHub on 2026-10-27; it was previously distributed to GRobot members only. The build did not record its source commit; tag qgc-kevinjiang-v2.0.0 points at d736dd1fe, the commit that recorded this packaging and its release notes.

This is not an official QGroundControl release. The project retains upstream license and provenance information.

QGC_KevinJiang v1.5.0

QGC_KevinJiang v1.5.0

Release date: 2026-10-05

QGC_KevinJiang v1.5.0 is an internal Windows release for GRobot project members, based on QGroundControl v5.0.8. It retains the stable QGC_KevinJiang application identity and supports upgrading an existing installation while retaining user settings.

Changes

• Add RTSP transport selection to the DeepShark four-camera panel: automatic (from the URL), TCP, or UDP. The setting applies to all RTSP channels and reconnects active video streams when saved, without editing individual URLs.
• Fix GStreamer receiver reference leaks, incomplete shutdown after clearing a URL, cross-thread state races, and video-size extraction.
• Move AI settings, process management, and detection overlays into the DeepShark custom plugin. Display detection overlays in the DeepShark video panel and retain compatibility with the previous overlay setting.
• Extend focused regression coverage for AI detection reception, RTSP transport selection, and video receiver lifecycle.
• Pin the event library to a revision compatible with the existing MAVLink dependency, preventing clean-build failures caused by updates to its floating branch.

Installation and upgrades

Use QGC_v1.5.0_KevinJiang-installer.exe. Close QGC before running the installer. The installer retains the fixed application identity and invokes the previous uninstaller with user-data retention enabled. Manual removal of an earlier version is not required.

After installation, confirm version 1.5.0 under Application Settings → Help. The Help page reads the runtime application version; the standard release script synchronizes the application and installer versions.

Release verification

This release passed audit run 20261005-182513. The final status is succeeded, with all six steps passing:

• Both installer and application versions are 1.5.0.
• Required Qt and GStreamer runtime files, staged application startup, the fixed installation identity, and the user-data retention upgrade contract were verified.
• Installer size: 168846774 bytes (approximately 161.02 MiB).
• Installer SHA-256: CDFD0E569D0103B71D0BD1A80E402AC781E6DD933772D221FC8156A66E9C156A.
• Authenticode status: not signed.

Following a manual upgrade from v1.4.0 on the local machine, the installation record and installed application were verified as 1.5.0, with one QGC_KevinJiang installation retained. The user confirmed the Help page version. Detailed logs are available under release-audit/20261005-182513/ in the distribution directory.

Known limitations

• Receiver defect fixes do not establish that intermittent field stutter has been resolved. Long-running, multi-stream RTSP operation still requires field validation.
• Python, CUDA dependencies, and AI model files are not bundled; configure them separately on each machine.
• RTSP URLs, gripper PWM limits, and joystick mappings remain device-specific.
• The release audit checks staged application startup and the upgrade contract; it does not replace real-device acceptance or an actual in-place installation test.
• The current workflow does not apply an Authenticode signature.

Provenance

This is an internal GRobot distribution, not an official QGroundControl release. The project retains upstream Git history and license information.

Source

Published on GitHub on 2026-10-27; it was previously distributed to GRobot members only. The build did not record its source commit; tag qgc-kevinjiang-v1.5.0 points at db9c91596 (the 1.5.0 checkpoint, also backup/pre-qgc-v5.1.5), the closest commit by packaging time and version.

QGC_KevinJiang v1.4.0

QGC_KevinJiang v1.4.0

Release date: 2026-09-09

QGC_KevinJiang v1.4.0 is an internal Windows release for GRobot project members, based on QGroundControl v5.0.8.

Highlights

• Improved multi-stream video connection, restart, and recovery workflows, including recovery handling for individual channels after reconnection.
• Added optional DeepShark connection monitoring after a stable baseline is established. It detects lost flight-controller heartbeat and enabled video streams that stop receiving new frames.
• Added an on-demand status banner, event and recovery records, configurable thresholds, and local alert tones to support operator inspection.
• Fixed lifecycle races exposed by disconnection scenarios and added focused regression coverage for connection monitoring, video control, camera management, and vehicle-link handling.

Installation and upgrade

Close QGC before running the installer. Existing QGC_KevinJiang installations upgrade in place and retain user data; manual uninstall is not required.

Verified artifact

• Installer and application version: 1.4.0
• SHA-256: 63A2CE519AB246497ED1920D0BA9A3AF82A9A397D1F7C0F8E2B9BF6F43C91684
• Authenticode status: not signed

Important boundary

Connection alerts are local operational prompts only. They do not diagnose the cause of a disconnection, change flight-controller parameters, or issue control commands.

QGC_KevinJiang v1.2.0

QGC_KevinJiang v1.2.0

Release date: 2026-08-10

QGC_KevinJiang v1.2.0 is an internal GRobot update based on QGroundControl v5.0.8. It retains the same application identity as the earlier QGC_KevinJiang releases, so the Windows installer upgrades the existing installation rather than installing a second copy.

Update Summary

• Added DeepShark video workspace enhancements, including a panoramic layout and drag-to-reorder video tiles.
• Added AUV mission and debug presentation workflows, including mission path display, task controls, debug speed, and emergency-stop confirmation.
• Added a 3D attitude workspace and a compact side view that can switch between RTSP channel 1 and the 3D view.
• Optimized the ground-control-to-vehicle connection path and improved connection error handling.
• Added compatibility improvements for newer ArduSub firmware and resolved related setup, parameter, and runtime error reports.
• Improved vehicle setup, sensor diagnostics, thruster mapping layout, and ArduSub arming checks.
• Updated application and Windows file version metadata to 1.2.0.
• Preserved the NSIS application identity, install directory, shortcuts, and uninstall key as QGC_KevinJiang.

Installation and Upgrade

Install using the release asset:

• QGC_v1.2.0_KevinJiang-installer.exe

If an earlier QGC_KevinJiang version is already installed, close it first, then run the v1.2.0 installer. It uses the same Windows installation identity and should replace the installed program files in place while retaining user configuration.

Known Limitations

• AI dependencies and model files are not included in the installer; operators configure local Python and model paths when AI detection is required.
• RTSP camera URLs are site-specific and need to be configured for the target equipment.
• This is an internal GRobot distribution, not an official QGroundControl release.

Provenance

QGC_KevinJiang v1.2.0 is based on QGroundControl v5.0.8. The upstream Git history remains in this repository for traceability and license compliance.

QGC_KevinJiang v1.1.1

QGC_KevinJiang v1.1.1

Release date: 2026-07-08

QGC_KevinJiang v1.1.1 is an internal GRobot maintenance update based on QGroundControl v5.0.8. It keeps the same application identity as v1.0.0 and v1.1.0, so the Windows installer upgrades the existing QGC_KevinJiang installation instead of installing a second copy.

Update Summary

• Updated application and Windows file version metadata to 1.1.1.
• Preserved the NSIS application identity, install directory, shortcuts, and uninstall key as QGC_KevinJiang.
• Preserved upgrade behavior from earlier QGC_KevinJiang versions while leaving user data in place.
• Kept optional AI dependencies and model files outside the installer.

Installation

Install using the GitHub Release asset:

• QGC_v1.1.1_KevinJiang-installer.exe

If an earlier QGC_KevinJiang version is already installed, close QGC first, then run the v1.1.1 installer. The installer should upgrade the existing QGC_KevinJiang installation and preserve user configuration.

Known Limitations

• AI dependencies and model files are not bundled with the installer.
• RTSP camera URLs still need to be configured by each operator.
• This is an internal GRobot distribution and is not an official QGroundControl release.

Provenance

QGC_KevinJiang v1.1.1 is based on QGroundControl v5.0.8. The upstream Git history is preserved for source provenance, licensing, and future traceability.

QGC_KevinJiang v1.0.0

QGC_KevinJiang v1.0.0

Release date: 2026-06-01

QGC_KevinJiang v1.0.0 is the first stable internal GRobot release. It is based on QGroundControl v5.0.8 and adds DeepShark video workflow support, optional local AI detection integration, and Windows installer packaging.

New Features

• Added DeepShark four-channel RTSP video panel.
• Added configurable camera names and RTSP URLs for front, left, right, and rear views.
• Added DeepShark status panel with per-camera connection state and retry controls.
• Added optional YOLO AI detection overlay.
• Added built-in AI settings, environment check, and AI start/stop/restart controls.
• Added GRobot application icon and Windows installer branding.
• Added Help page attribution for the QGC_KevinJiang build.

Fixes And Packaging Changes

• Fixed RTSP AI detection startup for camera URLs such as .sdp?real_stream by routing stream URLs through an Ultralytics .streams input file.
• Removed old built-in AI test RTSP defaults; installer defaults leave RTSP URLs empty.
• Packaged lightweight AI bridge scripts into the Windows install directory under bin/ai_detection.
• Kept large Python dependencies and YOLO model files outside the installer.
• Disabled QGC application update prompts for this internal build.
• Prevented native QGC video settings from automatically switching from disabled to UDP h.264 after vehicle connection.

Installation

Install using the GitHub Release asset:

• QGC_v1.0.0_KevinJiang-installer.exe

Do not commit the installer into the source repository. It should remain a Release attachment.

Optional AI Setup

AI detection requires a local Python environment with:

• ultralytics
• torch

The user also needs a local YOLO model file, for example:

• yolov8n.pt
• a custom trained .pt model

Configure the Python executable and model path in QGC Video settings, then run the environment check before starting AI.

Known Limitations

• AI dependencies and model files are not bundled with the installer.
• AI performance depends on the user's local Python, PyTorch, CUDA, GPU, and network camera environment.
• RTSP camera discovery is not automatic in v1.0.0; camera URLs must be entered manually.
• This is an internal GRobot distribution and is not an official QGroundControl release.

Provenance

QGC_KevinJiang v1.0.0 is based on QGroundControl v5.0.8. The upstream Git history is preserved to keep source provenance, licensing, and future traceability clear.