July 20, 2026 PX4 Weekly Dev Update: v1.18 Enters Beta, SIH Simulation Transition, and Enhanced Core Safety Features

PX4 Weekly Developer Briefing

Last week, the PX4 Autopilot project saw active development and significant changes across multiple fronts. Most notably, the official release of v1.18.0-beta1 signals the earnest beginning of stabilization efforts for the upcoming major version. In particular, the integrated testing environment is undergoing a major transition from Gazebo Classic to PX4’s own SIH (Software-In-the-Loop Hardware) simulation, which is expected to significantly enhance the efficiency and deterministic nature of the development and testing processes.

GitHub activity highlighted efforts to bolster system stability and safety. Numerous additions of mathlib related unit tests contributed to the robustness of core libraries, while bug fixes and feature enhancements were implemented across various subsystems including MAVLink, UAVCAN, and GPS drivers. Furthermore, ADS-B-based traffic avoidance and flight control related safety features are continuously being strengthened, marking a significant advancement in elevating drone safety in complex operational environments.

Community forums revealed keen interest in forward-looking technology integration, such as discussions on developing an AI-based flight diagnostic assistant within QGroundControl and exploring the potential of cuVSLAM as a Visual Inertial Odometry (VIO) system. Overall, PX4 demonstrates a balanced development trajectory, pursuing stabilization of core functionalities while embracing the latest technological trends and exploring ecosystem expansion.

PX4 Releases and Key Version-Specific Changes

This week, the PX4 Autopilot project officially kicked off the transition to its next major version with the release of the first beta for v1.18.

• v1.18.0-beta1 (Released on July 08, 2026)
  • Features: This is a pre-release for flight testing, encompassing the core changes of v1.18. Detailed release notes are available in the official documentation.
  • Key Changes:
    • The transition to an integrated testing environment based on internal SIH (Software-In-the-Loop Hardware) simulation has officially begun. (See GitHub Updates section for more details)
    • Includes stabilization and bug fixes for MAVLink and other subsystems.
    • Various driver and control logic improvements have been made.
  • Background: This beta release is a crucial step for developers to test new features, identify potential issues early, and provide feedback for the upcoming v1.18 stable version.

Core GitHub Updates (PX4-Autopilot)

Over the past 7 days, the PX4-Autopilot GitHub repository saw significant activity focused on improving code quality, critical infrastructure changes, and enhancing safety features.

Key Merged Pull Requests

This week was marked by improvements in testing infrastructure, bug fixes, and the addition of some new features.

• feat(tests): run integration tests with PX4 internal SIH simulation (Closed)
  • Analysis: The transition of the integrated testing environment from Gazebo Classic to PX4’s own SIH simulation is the most significant change this week. This will greatly contribute to enhancing the deterministic nature of tests, streamlining development workflows, and ultimately improving code quality and stability. Multiple commits and labels (risk:safety-critical, scope:simulation, scope:infrastructure, etc.) underscore the importance of this PR.
• ci(ros_integration_tests): migrate from Gazebo Classic to SIH (Closed)
  • Analysis: In conjunction with the above PR, ROS integration tests have also been migrated to SIH-based simulation. This impacts ROS users, who will need to understand the new simulation environment.
• feat(crsf): add local altitude/vertical-speed telemetry and fix GPS frame (Closed)
  • Analysis: Local altitude and vertical speed telemetry features have been added to the CRSF (Crossfire) driver, enhancing information exchange capabilities with remote control systems. The GPS frame fix contributes to data accuracy.
• fix(uavcan): publish non-valid range readings with quality 0 (Closed)
  • Analysis: The UAVCAN driver has been modified to publish invalid range readings with a quality of 0. This is a critical bug fix that improves sensor data reliability and prevents system malfunctions due to erroneous data.
• feat(adsb): improve traffic warnings, and warn only when disarmed (Closed)
  • Analysis: The ADS-B traffic warning system has been improved, with warning logic changed to only trigger when disarmed. This reduces unnecessary warnings during operation and allows for greater focus on actual threat situations. The risk:safety-critical label underscores the importance of this feature.
• feat(gps): inject SPARTN corrections alongside RTCM (Closed)
  • Analysis: SPARTN correction data injection functionality has been added to the GPS driver. This supports the SPARTN protocol alongside RTCM, improving RTK/PPP (Real-Time Kinematic/Precise Point Positioning) accuracy and ensuring compatibility with a wider range of GNSS correction services.
• fix(gps): split RTCM corrections and moving-baseline uORB topics (Closed)
  • Analysis: GPS RTCM correction data and moving-baseline uORB topics have been split, resolving potential issues in dual-GPS and fixed base station setups. This enhances the stability of precise positioning systems.
• Numerous test(mathlib) PRs (sq helper, float and int coverage, three-argument min and max, geo helpers, piecewise interpolateN, FilteredDerivative, isFinite): Significantly enhanced code robustness by extensively expanding the unit test coverage of the core mathlib.
• Numerous documentation and logging typo fixes (e.g., GPS config, EKF2 param, MC Pos Control, Zenoh docs): Represent efforts to improve overall documentation and code readability.

Open Critical Issues and Pull Requests

Several critical issues and PRs are currently awaiting resolution.

• fix(PositionControl): Prevent vehicle flipping on landing if MPC_ACC_DECOUPLE disabled (Open)
  • Analysis: This PR addresses a severe bug that causes the vehicle to flip on landing when the MPC_ACC_DECOUPLE parameter is disabled. The risk:safety-critical label indicates that it requires prioritized review and merging.
• [Bug] fmu-v6xrt: hardfault in the MAVLink Onboard-stream / uORB Subscription::update path under load (Open)
  • Analysis: A hardfault bug occurs in the MAVLink Onboard-stream / uORB Subscription::update path under high load on FMU-v6xrt boards. This can critically impact system stability and requires in-depth investigation.
• feat(ekf2): Model-based 3D wind estimator for fixed-wing (Open)
  • Analysis: This PR adds a model-based 3D wind estimator for fixed-wing aircraft. It has the potential to significantly improve the accuracy and stability of fixed-wing flight and is an essential component for advanced autonomous flight. 25 comments indicate active discussion.
• feat(gimbal): support mount gimbal manager (Open)
  • Analysis: This PR adds support for a gimbal mount manager. It standardizes the integration and control of camera or sensor gimbals mounted on drones, a crucial feature for enhancing payload operationality.
• feat(commander): add failsafe for traffic avoidance system (Open)
  • Analysis: This PR adds a failsafe feature for the traffic avoidance system. Linked with ADS-B related functionalities, it provides a vital safety mechanism to reduce the risk of collisions with other aircraft during flight.
• feat(boards): add Lectron Pi5 Autopilot board (Open)
  • Analysis: This PR adds support for the new Lectron Pi5 Autopilot board, showcasing the expandability of hardware platforms supported by PX4.

Weekly Dev Call & Community Trends

Last week, the PX4 Discourse forum hosted various technical discussions alongside the regular developer call.

• PX4 Dev Call: Jul 15, 2026 (Team sync, and Community Q&A) (링크)
  • Analysis: The regular weekly developer meeting was held, featuring team synchronization and a community Q&A session. Although there were no forum comments, it is anticipated that internal coordination among the development team and direct responses to community questions took place.
• PX4 Autopilot Section Key Discussions:
  • Contributing atxxxx analog OSD support (링크)
    • Analysis: This discussion concerns contributing support for a specific analog OSD (On-Screen Display). It demonstrates PX4’s continuous pursuit of integration with various external hardware.
  • How to tune cascaded PID gains for a custom PX4 SITL drone? (링크)
    • Analysis: This is a question about tuning cascaded PID gains for a custom PX4 SITL drone, representing a request for support on common technical challenges faced by new developers or custom drone builders.
  • cuVSLAM as a medium for Visual Inertial Odometry (VIO) (링크)
    • Analysis: The discussion on the feasibility of implementing VIO using cuVSLAM indicates the PX4 community’s strong interest in integrating advanced perception and navigation technologies beyond mere flight control. This is part of a broader movement to expand the capabilities of autonomous drones through the combination of high-performance computing and sensor fusion technologies.
  • MicroXRCEAgent isn’t transmitting ROS2 topics (링크)
    • Analysis: This discussion addresses issues with ROS2 topic transmission. Integration between PX4 and ROS2 is crucial for many developers, but it highlights ongoing challenges in setup and debugging.

Subsystem Trends (MAVLink, MAVSDK, QGC)

Significant changes and discussions have also been observed within the key subsystems of the PX4 ecosystem.

MAVLink Trends

• GitHub PR: fix(mavlink): select GPS_RAW_INT/GPS2_RAW instance via SENS_GPS_PRIME (Open) is an ongoing critical bug fix to correctly report data from the second GPS sensor via MAVLink messages. This directly impacts multi-GPS system users.
• GitHub PRs: fix(mavlink): use published Open Drone ID Basic ID (Closed) and fix(mavlink): handle unknown battery time remaining (Closed) demonstrate efforts to enhance the accuracy of information transmission and adherence to standards via the MAVLink protocol.
• GitHub PR: fix(mavlink_command): allow QGC 5 takeoff command (again) (Open) attempts to resolve MAVLink command compatibility issues with QGroundControl 5. This is essential for seamless interaction with GCS.
• Analysis: MAVLink continues to be refined in its handling of various messages and commands, with multi-sensor support and maintaining compatibility with external GCS appearing as crucial development directions.

MAVSDK Trends

• GitHub PR: feat(tests): add SIH as a selectable MAVSDK test backend (Closed) enables the use of PX4’s new SIH simulation backend for MAVSDK integration tests. This shows MAVSDK aligning with PX4 core’s testing environment changes, providing a more stable and predictable testing environment for application development with MAVSDK.
• GitHub PRs: docs(zh/ko/uk): point MAVSDK SITL tests at sih-sitl.json (Closed for zh, Open for ko, Open for uk), among others, are documentation updates stemming from the SIH transition, helping developers worldwide easily understand and utilize the new testing environment.
• Analysis: MAVSDK is responsive to changes in the PX4 core, striving to provide a more robust development experience for external application developers, particularly through improvements in its testing infrastructure.

QGroundControl (QGC) Trends

• Discourse: Building a Safety-Bounded AI Flight Diagnostic Assistant in QGroundControl (링크)
  • Analysis: Active discussions are underway regarding the development of a safety-bounded AI flight diagnostic assistant within QGC. This represents a significant effort for QGC to evolve beyond a mere ground control station into an intelligent system that analyzes safety-related data in real-time during flight and provides meaningful information to the pilot. Such capabilities hold substantial potential to greatly enhance the reliability and operational efficiency of autonomous drones.
• Discourse: MAVLab: an Android app that presents to QGroundControl as an ArduPilot-like vehicle (open-source teaching simulator) (링크)
  • Analysis: This discussion revolves around an interesting project utilizing QGC as an interface for an external simulator. It demonstrates QGC’s flexibility for use in various educational and research environments.
• Analysis: QGC is evolving beyond user interface and usability, pursuing the integration of cutting-edge technologies like AI to enhance drone operational safety and intelligence. Furthermore, its role as an extensible platform continues to be strengthened.