UoSM UAV Stack

A complete PX4-based autonomous UAV stack using a stereo camera (ZEDM) and 2D LiDAR (RPLidar S1) for under-canopy surveying in oil palm plantations. Virtual map generation and global path planning are available at PalmOilDetectree2.

Prerequisites

• ROS 2 Humble
• Jetpack 6.2 / Ubuntu 22.04 LTS

A. Build

Install dependencies:

chmod +x ./scripts/setup.sh
./scripts/setup.sh

Build workspace:

source /opt/ros/humble/setup.bash
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release --parallel-workers $(nproc)

Optional: experimental NMPC setup

# 1. Build acados vendor package
colcon build --packages-select acados_vendor_ros2
source install/setup.bash

# 2. Patch acados_vendor for Jetson (aarch64 t_renderer + Python 3.10 fix)
./src/nmpc_controller/scripts/patch_acados_vendor.sh

# 3. Ensure compatible numpy and scipy
pip3 install numpy==1.26.4 scipy==1.15.3

# 4. Generate the acados solver C code
cd src/nmpc_controller/acados_model
python3 generate_solver.py
cd ../../../

# 5. Build the NMPC controller
colcon build --packages-select nmpc_controller

B. Run the Simulation

For interactive mode, once RViz opens use the 2D Nav Goal tool to click waypoints. The ego planner will generate a collision-free trajectory and the controller will track it.

For preset waypoint mode, the planner will begin executing the loaded waypoints automatically after takeoff, waypoints and map currently in maps.

Terminal 1 — Start PX4 SITL

cd ~/PX4-Autopilot && HEADLESS=1 make px4_sitl gz_x500

Wait until PX4 finishes initialization and launch QGC.

Terminal 2 — Launch the simulation

Two launch files are available depending on how you want to supply waypoints:

Interactive mode — click waypoints in RViz using the 2D Nav Goal tool:

source install/setup.bash
ros2 launch uosm_uav_bringup sim_interactive_single.launch.py

Preset waypoint mode — follows a pre-defined waypoint CSV file:

source install/setup.bash
ros2 launch uosm_uav_bringup sim_preset_wp_single.launch.py

C. Deployment

Reboot PX4 Autopilot to reset internal state

source install/setup.bash
ros2 launch uosm_uav_bringup reboot_px4.launch.py

Sanity check for hardware and PX4 tuning (hover stability, vibration metrics, EKF2 health, PID behavior, battery). This helps confirm stable position-mode flight without GNSS.

source install/setup.bash
ros2 launch uosm_uav_bringup real_sanity_check.launch.py use_rosbag:=false

IMPORTANT: Real flight, make sure to fly in safe area. Preflight checklist:

• Verify RC kill switch and failsafe behavior.
• Confirm battery level above arm threshold.
• Verify EKF2 and local position estimate are healthy. ReCalibrate if needed.
• Confirm long range telemetry working to maintainn ground control link.
• Confirm wifi/network static ip and dds setup correctly.

KNOWN ISSUE: For rosbag recording, avoid recording too many heavy topics; File I/O operation can significantly impact flight performance.

source install/setup.bash
# ros2 launch uosm_uav_bringup real_jetson_orin_interactive.launch.py use_foxglove:=true
ros2 launch uosm_uav_bringup real_jetson_orin_preset_wp.launch.py use_foxglove:=true wait_for_alignment:=true use_rosbag:=true

ros2 launch uosm_uav_bringup real_minimal.launch.py use_foxglove:=true

Repository Overview

Root layout

Path	Purpose
src/	Main ROS 2 packages
scripts/	Setup and helper scripts
bags/	Recorded rosbag data (real and simulation)
px4_related/	PX4 params and related artifacts
jetson_config/	Jetson device-specific configs
build/, install/	Colcon build outputs

src/ package structure

Group	Main packages	Purpose
Core bringup and interfaces	uosm_uav_bringup, uosm_uav_interface, uosm_robot_viewer	System launch/config, custom interfaces, robot model visualization
Planning and control	nmpc_controller, path_planning_module/grid_map, path_planning_module/planner_manager, path_planning_module/planner_utils	NMPC trajectory tracking, occupancy mapping, planning, trajectory utilities
Perception	perception_module/odom_republisher, perception_module/map_alignment, perception_module/trunk_segmentation, perception_module/jetson_gscam2, perception_module/sllidar_ros2, perception_module/zed-ros2-wrapper	Odometry fusion/alignment, sensing, camera/lidar integration
Virtual and simulation support	virtual_module/local_sensing, virtual_module/map_processor, virtual_module/odom_visualization, virtual_module/pose_utils	Simulated sensing, map loading, odometry display, pose/transform helpers
Third-party vendors/tools	third_party/acados_vendor_ros2, third_party/ros2_jetson_stats	External solver and Jetson monitoring integration

License

This project is licensed under the GNU General Public License v3.0 (GPL-3.0). See LICENSE for full terms.

Citation???

Cite so that my supervisor can keep his job and get a promotion.

@article{uosm_uav_stack,
  title   = {xxx},
  author  = {Janitor},
  year    = {2026},
  url     = {xxx},
  note    = {xxx}
}

Contact

• Email: skl1g14@soton.ac.uk / lsk950329@hotmail.com