Skip to content

Software

Scope of this document

This document covers low-level software only — RPi system configuration, ROS2 stack, VIO, PX4, and peripheral firmware. High-level autonomy and computer vision are covered in separate sections below, but not documented in detail here.

1. Low-level software

1.1 Raspberry Pi — System configuration

  • OS: Ubuntu 24.04
  • Hostname: Drone26-U24
  • ROS2 distribution: Jazzy (native, no LXD container)
  • ROS2 sourced automatically in ~/.bashrc:
    source /opt/ros/jazzy/setup.bash
  • Workspace: ~/ros2_ws
  • Workspace sourced automatically in ~/.bashrc:
    source ~/ros2_ws/install/setup.bash

Network configuration

Interface Mode IP / Notes
eth0 Static 10.10.10.10 / 255.255.255.0
wlan0 DHCP Connects to configured WiFi networks

Static IP configured via NetworkManager. To add a new WiFi network:

sudo nmcli dev wifi connect "NETWORK_NAME" password "PASSWORD"

Startup scripts

Two scripts in ~/ manage the drone session:

~/start_drone.sh   # starts tmux session with all processes
~/stop_drone.sh    # sends Ctrl+C to all windows and kills session

See general.md for the full startup procedure.

1.2 MicroXRCE-DDS Agent

Bridge between PX4 and ROS2. Runs on the RPi, communicates with Pixhawk via UART.

MicroXRCEAgent serial --dev /dev/ttyAMA0 -b 921600

Must be started before Pixhawk finishes booting

See general.md — startup procedure section.

Installation:

# [TO FILL] — document installation steps used

1.3 RealSense D435i driver

Package: realsense2_camera (ROS2)

Launch command used in start_drone.sh:

ros2 launch realsense2_camera rs_launch.py \
  enable_gyro:=true \
  enable_accel:=true \
  unite_imu_method:=1 \
  rgb_camera.color_profile:=640x480x30 \
  enable_depth:=false
  • Depth disabled — not used by OpenVINS, saves USB 2.0 bandwidth
  • IMU topics unified via unite_imu_method:=1

Topics published:

Topic Content
/camera/camera/color/image_raw RGB image 640x480 @ 30fps
/camera/camera/imu Unified IMU data

Installation:

# [TO FILL] — document installation steps used

1.4 Camera-IMU calibration (Kalibr)

Calibration was performed with Kalibr using an AprilGrid target printed in A3 format.

Recalibrate after any camera or IMU mount change

If the RealSense is remounted or its orientation changes, calibration must be redone.

Calibration is a f***ing pain in the ass

Please do it one time only, be sure about everything else, than calibrate.

AprilGrid target parameters

Parameter Value
Tag size [TO FILL] m
Tag spacing [TO FILL]
Rows [TO FILL]
Cols [TO FILL]

Calibration procedure

1 — Record calibration bag

ros2 bag record --topics /camera/camera/color/image_raw /camera/camera/imu -o ~/calibration_bag

Move the drone slowly in front of the AprilGrid for ~90 seconds, covering all angles.

2 — Convert bag (mcap → ROS1 format)

# [TO FILL] — document conversion steps used

3 — Run Kalibr

# [TO FILL] — document Kalibr command used

4 — Output files

Kalibr produces camchain.yaml and imu.yaml. Copy the relevant values into the OpenVINS config.

Current calibration values

# [TO FILL] — paste current camchain.yaml and imu.yaml values here after recalibration

1.5 OpenVINS

Visual-Inertial Odometry system. Fuses RealSense RGB + IMU data to estimate drone position.

Package: ov_msckf Config folder: ~/ros2_ws/src/open_vins/config/d435i/

Launch command used in start_drone.sh:

ros2 launch ov_msckf subscribe.launch.py \
  config:=d435i \
  max_cameras:=1 \
  use_stereo:=false

Key config file: ~/ros2_ws/src/open_vins/config/d435i/estimator_config.yaml

ZUPT parameters

Zero velocity update — keeps position stable when the drone is stationary.

try_zupt: true
zupt_chi2_multipler: [TO FILL]
zupt_max_velocity: [TO FILL]
zupt_noise_multiplier: [TO FILL]
zupt_max_disparity: [TO FILL]
zupt_only_at_beginning: false

Camera-IMU time offset

calib_cam_timeoffset: false   # online calibration disabled — use fixed value from Kalibr
# [TO FILL] — add fixed timeoffset value

Topic published:

Topic Content
/ov_msckf/odomimu Position + orientation estimate

Known issue — VIO drift

OpenVINS currently drifts significantly during flight, especially on low-texture surfaces (grass). Position mode is unusable until this is resolved. Current config values may be incorrect — recalibrate and retune before relying on VIO for autonomous flight.

Installation:

# [TO FILL] — document build steps used

1.6 VIO Bridge

Custom Python script that subscribes to OpenVINS odometry and publishes it to PX4 via MicroXRCE-DDS, so PX4 EKF2 can use VIO as a position source.

Location: ~/ros2_ws/vio_bridge.py

Run:

source ~/ros2_ws/install/setup.bash && python3 ~/ros2_ws/vio_bridge.py

How it works

OpenVINS → /ov_msckf/odomimu → vio_bridge.py → /fmu/in/vehicle_visual_odometry → MicroXRCE-DDS → PX4 EKF2

The script:

  1. Subscribes to /ov_msckf/odomimu (ROS2 Odometry message)
  2. Converts position and orientation to the PX4 NED frame
  3. Publishes to /fmu/in/vehicle_visual_odometry (PX4 uXRCE-DDS topic)

1.7 PX4 parameters

Parameters configured on the Pixhawk 6C for this setup. Set via QGroundControl → Parameters.

uXRCE-DDS (ROS2 bridge)

Parameter Value Notes
UXRCE_DDS_CFG TELEM2 UART bridge to RPi
SER_TEL2_BAUD 921600 Must match MicroXRCE agent
UXRCE_DDS_DOM_ID 0 Default domain

Telemetry (ESP8266)

Parameter Value Notes
SER_TEL1_BAUD 57600 Must match ESP8266 firmware config
MAV_0_MODE Normal
MAV_0_RATE [TO FILL]

RC / SBUS

Parameter Value Notes
[TO FILL] [TO FILL] SBUS configuration

EKF2 — VIO fusion

Parameter Value Notes
[TO FILL] [TO FILL] EKF2 VIO parameters

GPS - OFF

Parameter Value Notes
[TO FILL] [TO FILL] GPS parameters

ESC / Motors

Parameter Value Notes
[TO FILL] [TO FILL] DShot600 configuration

VL53L1X range sensor

Parameter Value Notes
[TO FILL] [TO FILL] Range sensor configuration

1.8 ESP8266 — MavESP8266 firmware

  • Firmware: MavESP8266 (MAVLink WiFi bridge)
  • Web interface: 192.168.4.1 (connect to TechDrone_Telem first)
  • Baud rate: 57600 (must match SER_TEL1_BAUD in PX4)
  • UDP port: 14550

Flashing:

# [TO FILL] — document flashing procedure

1.9 ELRS firmware

  • Protocol on drone side: SBUS (output from receiver to Pixhawk)
  • Binding: via passphrase — already configured, stable
  • Firmware version: [TO FILL]

Rebinding procedure:

# [TO FILL] — document rebinding steps if needed

2. High-level software / Autonomy

Not documented here

The autonomy stack is developed in Matlab/Simulink and runs on the Raspberry Pi. It publishes commands via ROS2 / MicroXRCE-DDS topics to PX4. Documentation for this layer is the responsibility of the autonomy sub-team.

From the low-level perspective, the autonomy stack is just another ROS2 node publishing to PX4 topics — no special configuration required at this level.

3. Computer Vision

3.1 ArUco detection

Work in progress — not documented yet

To be completed by the computer vision sub-team.

3.2 Probe detection

Work in progress — not documented yet

To be completed by the computer vision sub-team.

TODO

  • Fill in all [TO FILL] placeholders — requires drone in front of you
  • Document complete RPi setup from scratch (OS install, ROS2, all packages) — essential for filesystem recovery
  • Fix OpenVINS drift — current config and calibration values are unreliable
  • Recalibrate camera-IMU with Kalibr after fixing camera mount definitively
  • Document full PX4 parameter list (export from QGC and paste here)

Low-level / RPi maintainer: Alessandro Paolantonio — High-level autonomy / MATLAB Simulink maintainer: Samuele Passeri