How to modify drone configuration (sensors, algorithms, control logic) in this project?

[dynamics986]

A promising project! I'm working with it and trying to understand the best practices for customizing drone configurations. I want to develop a system for multi-drone collaboration so that every drone can detect other drones’ location by sensor. Is it possible to develop starting from this project?

I read the code but have not found where to modify the property of different drones and how to define my own drone and use them. For example, customize perception algorithms (e.g., swapping YOLO models, modifying detection pipelines) and tune control/logic parameters (e.g., adjusting PID gains, modifying flight behaviors, changing mission planning logic).

[JacopoPan]

Hello @dynamics986

how to define my own drone

All models are in folder aircraft_models and instantiated with PX4/ArduPilot SITL in simulation.yml.erb

aerial-autonomy-stack/simulation/simulation.yml.erb

Lines 77 to 92 in e67cd92

	<% drones.each_with_index do |drone, sitl_instance| %>
	<%
	drone_id = sitl_instance + 1
	is_quad = drone[:type] == 'quad'
	if autopilot == 'px4'
	autostart = is_quad ? 5140 : 5141
	model_name = is_quad ? "x500_#{sitl_instance}" : "standard_vtol_#{sitl_instance}"
	x_offset = drone[:index] * 2
	y_offset = is_quad ? x_offset : x_offset + 2
	# Env var parameter injection (using prefix PX4_PARAM_*) of https://docs.px4.io/v1.17/en/advanced_config/parameter_reference#UXRCE_DDS_AG_IP
	signed_32bit_ip = "#{sim_subnet}.90.#{drone_id}".split('.').map(&:to_i).pack('C*').unpack1('l>')
	elsif autopilot == 'ardupilot'
	vehicle_config = is_quad ? '-v ArduCopter -f gazebo-iris' : '-v ArduPlane'
	model_name = is_quad ? "iris_with_ardupilot_#{drone_id}" : "alti_transition_quad_#{drone_id}"
	end
	%>

swapping YOLO models

Models are exported in the Dockerfile

aerial-autonomy-stack/tools_and_docs/docker/aircraft.dockerfile

Lines 331 to 339 in e67cd92

	WORKDIR /aas/yolo
	# Model options (from fastest to most accurate, <10MB to >100MB): yolo26n, yolo26s, yolo26m, yolo26l, yolo26x
	# Export standard 640 static as yolo26n_640.onnx and smaller 320 static as yolo26n_320.onnx
	RUN /yolo-env/bin/python3 -c "from ultralytics import YOLO; YOLO('yolo26n.pt').export(format='onnx', opset=12, imgsz=640)" && \
	mv yolo26n.onnx yolo26n_640.onnx && \
	/yolo-env/bin/python3 -c "from ultralytics import YOLO; YOLO('yolo26n.pt').export(format='onnx', opset=12, imgsz=320)" && \
	mv yolo26n.onnx yolo26n_320.onnx && \
	/yolo-env/bin/python3 -c "import json; from ultralytics import YOLO; print(json.dumps(YOLO('yolo26n.pt').names))" | grep '{' > coco.json && \
	rm yolo26n.pt

And used in yolo_py

aerial-autonomy-stack/aircraft/aircraft_ws/src/yolo_py/yolo_py/yolo_node.py

Lines 164 to 176 in e67cd92

	# Load YOLO model and runtime
	# Options, from fastest to most accurate, <10MB to >100MB: yolo26n, yolo26s, yolo26m, yolo26l, yolo26x, export in aircraft.dockerfile
	if self.architecture == 'x86_64':
	max_dim = max(stream_width, stream_height)
	if max_dim <= 320:
	print("Stream resolution is <=320. Selecting 320 model.")
	model_path = "/aas/yolo/yolo26n_320.onnx"
	self.input_size = 320 # YOLO input size
	else:
	print("Stream resolution is >320. Selecting 640 model.")
	model_path = "/aas/yolo/yolo26n_640.onnx"
	self.input_size = 640 # YOLO input size
	print("Loading CUDAExecutionProvider on AMD64 (x86) architecture.")

modifying detection pipelines

On Jetson, you can directly work on the CSI GStreamer pipeline with DeepStream

aerial-autonomy-stack/aircraft/aircraft_ws/src/yolo_py/yolo_py/yolo_node.py

Lines 113 to 124 in e67cd92

	else: # Default, acquire CSI camera
	# GPU pipeline:
	gst_pipeline_string = (
	f"nvarguscamerasrc sensor-id={self.camera_id} ! "
	"video/x-raw(memory:NVMM), width=1280, height=720, framerate=60/1 ! "
	"nvvidconv ! "
	"video/x-raw(memory:NVMM), format=RGBA ! "
	"nvdewarper config-file=/aas/aircraft_resources/patches/imx219_dewarper_config.txt ! "
	"nvvidconv ! "
	"video/x-raw, width=640, height=360, format=BGRx ! "
	"videoconvert ! video/x-raw, format=BGR ! "
	"appsink drop=true max-buffers=1 sync=false"

adjusting PID gains

PX4/ArduPilot parameters can be changed with the corresponding files in folder aircraft_models

aerial-autonomy-stack/simulation/simulation_resources/aircraft_models/x500/5140_gz_aas_x500

Lines 101 to 107 in e67cd92

	# Aggressive Position control
	# Tilt limits
	param set-default MPC_MAN_TILT_MAX 70.0
	param set-default MPC_TILTMAX_AIR 89.0
	# Acceleration limits
	param set-default MPC_ACC_HOR 10.0
	param set-default MPC_ACC_HOR_MAX 10.0

aerial-autonomy-stack/simulation/simulation_resources/aircraft_models/iris_with_ardupilot/ardupilot-4.6.params

Lines 5 to 8 in e67cd92

	ANGLE_MAX 3000
	PSC_ANGLE_MAX 0
	WPNAV_ACCEL 500
	WPNAV_ACCEL_Z 250

modifying flight behaviors

The ROS2 nodes that interact with PX4 and ArduPilot are:

• High-level ArduPilot: https://github.com/JacopoPan/aerial-autonomy-stack/blob/main/aircraft/aircraft_ws/src/autopilot_interface/src/ardupilot_interface.cpp
• High-level PX4: https://github.com/JacopoPan/aerial-autonomy-stack/blob/main/aircraft/aircraft_ws/src/autopilot_interface/src/px4_interface.cpp
• Low-level ArduPilot: https://github.com/JacopoPan/aerial-autonomy-stack/blob/main/aircraft/aircraft_ws/src/offboard_control/src/ardupilot_guided.cpp
• Low-level PX4: https://github.com/JacopoPan/aerial-autonomy-stack/blob/main/aircraft/aircraft_ws/src/offboard_control/src/px4_offboard.cpp

changing mission planning logic

Example missions are in folder missions and executed by the mission node