d2dtracker_interception

Interception strategy library for the drone-to-drone interception workspace. It holds pure-Python geometry and guidance functions (rendezvous point, Proportional Navigation) with no ROS dependency, plus two stand-alone nodes built on them. In the pluggable pipeline it is not launched as a node: interception_planning imports its functions for the rendezvous, pn_lead and head_on planner backends. The two nodes are separate "strategy" methods from RUNBOOK.md (Strategy A and Strategy B).

Build type: ament_python. License: MIT.

Library

Module	Functions	Used by
interception_point.py	reach_time (bang-bang accel-then-cruise time to a point), time_to_intercept, find_optimal_intersection (first predicted sample with reach_time <= k * dt; else the last sample)	interceptor node, planner rendezvous
interception_modes.py	interceptor_trajectory (Algorithm 1), closest_interception_sample (Algorithm 2), find_reference_sample (Algorithm 3), reference_trajectory (Algorithm 4, head-on), distance2D, dist3D, yaw_to_tgt	interceptor node, planner head_on
guidance.py	closing_velocity, los_rate, proportional_navigation (true PN, a = N * Vc * (omega x los_hat)), pn_velocity_command (LOS closing speed + PN lateral term, capped at vmax)	guidance node, planner pn_lead
frames.py	body_to_world, odom_velocity_world (rotate the MAVROS odom twist from body to world frame)	both nodes

States are 9-vectors [x, y, z, vx, vy, vz, ax, ay, az].

msg/State.msg and interception_pseudo_code are kept for reference; the message is not generated (an ament_python package cannot build interfaces).

Backends (strategies)

Name	Executable	What it does
Rendezvous (Strategy A)	interceptor	Publishes the rendezvous point and an acceleration-limited path toward it. Does not command the FCU
Proportional Navigation (Strategy B)	guidance	Flies the interceptor with PN velocity setpoints. Arms and requests OFFBOARD itself

Node: interceptor (node name interceptor_node)

Direction	Topic	Type	QoS
sub	interceptor_odom	nav_msgs/Odometry	sensor data
sub	target_predicted_path	nav_msgs/Path	depth 10
sub	target_odom	nav_msgs/Odometry	sensor data
pub	intercept_point	geometry_msgs/PoseStamped	depth 10
pub	interception_path	nav_msgs/Path	depth 10

Once a target_predicted_path message arrives its positions are used (velocity set to zero); otherwise a constant-velocity prediction is built from target_odom. Odom positions get the spawn offsets added; odom velocities are rotated from body to world frame.

Parameter	Default
vmax, amax	8.0, 4.0
prediction_dt	0.1 (s)
prediction_horizon	30 (samples, fallback prediction)
rate	10.0 (Hz)
frame_id	map
interceptor_spawn, target_spawn	[0.0, 0.0, 0.0], [10.0, 0.0, 0.0]
target_path_in_world	true (do not offset the predicted path)

capture_radius (1.0) is declared but not used by the loop.

Node: guidance (node name guidance_node)

Direction	Topic / service	Type	QoS
sub	mavros/state	mavros_msgs/State	reliable, transient local
sub	mavros/local_position/odom	nav_msgs/Odometry	sensor data
sub	target_odom	nav_msgs/Odometry	sensor data
pub	mavros/setpoint_raw/local	mavros_msgs/PositionTarget	sensor data
client	mavros/cmd/arming, mavros/set_mode	CommandBool, SetMode	-

Until it engages, it sends a position setpoint at its first x/y and takeoff_height. With climb_first it engages once a target odom has arrived, the vehicle is armed and altitude >= 0.9 * takeoff_height; with climb_first false it engages as soon as a target odom arrives. Engaged, it sends velocity setpoints with yaw along the velocity, using world-frame velocities of both drones: the PN law (law:=pn) or pure pursuit (law:=pursuit, the line of sight at closing_speed). With auto_arm, a 1 s timer requests OFFBOARD, then arming, once the loop has run 50 times. Reaching capture_radius only logs a message.

Parameter	Default
law	pn (pn or pursuit)
nav_gain	3.0 (PN constant N)
vmax	8.0
closing_speed	6.0
capture_radius	1.0
takeoff_height	4.0
auto_arm	true
climb_first	true
rate	30.0 (Hz)
interceptor_spawn, target_spawn	[0.0, 0.0, 0.0], [10.0, 0.0, 0.0]

Build

cd ~/drone_interception_ws/ros2_ws
colcon build --symlink-install --packages-select d2dtracker_interception
source install/setup.bash

Run

Every terminal needs the sim's RMW and domain:

export RMW_IMPLEMENTATION=rmw_fastrtps_cpp
export ROS_DOMAIN_ID=77   # must match the sim's ros_domain_id:= (default 77)

# Strategy A: rendezvous point + path
ros2 launch d2dtracker_interception interceptor.launch.py
# Strategy B: Proportional Navigation (flies the interceptor)
ros2 launch d2dtracker_interception guidance.launch.py nav_gain:=3.0 vmax:=8.0

• interceptor.launch.py: arguments interceptor_odom (/interceptor/mavros/local_position/odom), target_odom (/target/mavros/local_position/odom), target_predicted_path (/target/predicted_path), vmax, amax. The node has no namespace, so it publishes /intercept_point and /interception_path. Sets use_sim_time: True.
• guidance.launch.py: arguments namespace (interceptor), nav_gain, vmax, target_odom. Sets use_sim_time: True.

Run only one interceptor controller at a time. The guidance node must not run together with pipeline.launch.py heads fsm, offboard or none (their control_node publishes to the same setpoint topic) or with orchestrator:=bt.

In the pipeline, the library is selected through the planner:

ros2 launch drone_interception_sim pipeline.launch.py planner:=rendezvous
ros2 launch drone_interception_sim pipeline.launch.py planner:=pn_lead
ros2 launch drone_interception_sim pipeline.launch.py planner:=head_on

Tests

test/test_interception.py (7 tests), test/test_guidance.py (8 tests) and test/test_frames.py (3 tests) cover the library without a ROS runtime. test/ also has the copyright/flake8/pep257 lint tests.

cd ~/drone_interception_ws/ros2_ws/src/d2dtracker_interception
python3 -m pytest test

Measured on 30 Oct 2026: 18 passed, 1 skipped, 2 failed. The skip is the copyright test (marked skip). The 2 failures are lint, not functional: flake8 (I100 import order in setup.py) and pep257 (D213 docstring style).