A Progress-Aware Leader-Follower Midair Docking System for Dual-Drone Aerial Manipulation

The gist: A progress-aware leader-follower midair docking system for dual-drone aerial manipulation is presented, which utilizes a mission supervisor to manage phase transitions and enforce tolerance gates for repeatable engagement.

Platform and Hardware

The platform consists of two Crazyflie 2.1 microquadrotors operating in a leader-follower formation and docking using a lightweight modular frame with passive magnetic latching The platform integrates a complete hardware-software stack (ROS 2 with Crazyflie/PX4 interfaces) and synchronized logging for benchmark evaluation The modular docking frame is constructed from carbon-fiber rods and 3Dprinted connectors with embedded magnets, designed to minimize mass while maintaining sufficient stiffness for repeatable engagement

Control and Formation

The system employs a leader-follower formation model where the drones are assigned a fixed docking separation ddock along the real-world x-axis, with baseline vector, b = [ddock,0,0]⊤ The synchronized formation targets are then p⋆L = g−1/2b, p⋆F = g+1/2b (3) with fixed yaw references This opposing-yaw convention enforces a face-to-face docking geometry during approach The synchronized cruise speed is chosen conservatively from the two vehicle capabilities, vsync = 0.8min v(L) form,v(F) form (4) and the timeout is derived from the commanded formationcenter displacement tˆ = g−cvsync, tmax = clip(2tˆ,30 s,tusr) (5)

Progress-Aware Supervision

A progress-aware mission supervisor manages phase transitions: approach, alignment, capture, and settle This supervisor incorporates a fixed baseline with opposing yaw, synchronized setpoints, and progress gating It advances mission phases only when both drones meet specific distance, yaw, and velocity tolerances A phase transition is permitted only when eb < εb, eψ < εψ, vrel < εv (7) where εb, εψ, and εv are design tolerances

Evaluation and Results

Experiments in an indoor motion-capture arena and supporting simulation show a repeatable approach and docking behavior In simulation, the 3D trajectories converge to a stable docking geometry and constant-baseline formation maneuvers Real-world trials demonstrate that short capture transients (e.g., follower vz peaking near ∼ 2 m/s) decay within a few seconds, yielding bounded velocity profiles and stable engagement The framework demonstrates accurate relative tracking, robust stabilization under tight payload constraints, and repeatable midair docking behavior

Contributions

The main contributions of this work are i) A progressaware mission supervisor that formalizes phase gating for leader-follower docking using measurable guards, ii) A lightweight modular docking platform compatible with micro-UAV payload constraints, enabling repeatable midair engagement, iii) An integrated software pipeline (ROS 2 with Crazyflie/PX4 interfaces) with a synchronized measurement pipeline for dual-drone experiments, and iv) An experimental evaluation in simulation and real-world environments to quantify docking repeatability

Discussion

The results indicate that progress-aware phase gating and synchronized setpoint generation improve docking repeatability by reducing premature contact and limiting transient amplification during coupled leader-follower motion The lightweight passive docking interface further increases engagement robustness by absorbing small residual pose errors at contact while remaining compatible with micro-UAV payload constraints Future work will extend the platform in three directions: (i) introduce onboard relative sensing and time synchronization to reduce reliance on MoCap, (ii) expand benchmarking to a wider range of docking distances, approach speeds, and disturbance conditions, and (iii) port the supervision and evaluation pipeline to larger PX4-class UAVs

Conclusion

This paper presented a reliable dual-drone midair docking platform for cooperative aerial manipulation under tight micro-UAV payload and thrust constraints Two quadrotors operate in a leader-follower formation and execute midair docking using a lightweight modular frame with passive magnetic latching, supervised by a progress-aware mission supervisor that regulates phase transitions (approach, alignment, capture, and settle) and enforces tolerance-based guards to improve repeatability A complete hardwaresoftware stack (ROS 2 with Crazyflie/PX4 interfaces) was implemented, enabling synchronized estimation, setpoint generation, and systematic evaluation Simulation and indoor flight experiments demonstrate stable leader-follower docking behavior with bounded transients and consistent baseline/yaw regulation The benchmark-style metrics reported in this work (success rate, time-to-dock, baseline/yaw consistency, and failure-modes) provide a practical foundation for statistically grounded comparison of docking supervision and controller configurations A complete hardwaresoftware stack (ROS 2 with Crazyflie/PX4 interfaces) was implemented, enabling synchronized estimation, setpoint generation, and systematic evaluation The benchmark-style metrics reported in this work (success rate, time-to-dock, baseline/yaw consistency, and failure-modes) provide a practical foundation for statistically grounded comparison of docking supervision and controller configurations A complete hardwaresoftware stack (ROS 2 with Crazyflie/PX4 interfaces) was implemented, enabling synchronized estimation, setpoint generation, and systematic evaluation The benchmark-style metrics reported in this work (success rate, time-to-dock, baseline/yaw consistency, and failure-modes) provide a practical foundation for statistically grounded comparison of docking supervision and controller configurations The complete hardwaresoftware stack (ROS 2 with Crazyflie/PX4 interfaces) was implemented, enabling synchronized estimation, setpoint generation, and systematic evaluation The benchmark-style metrics reported in this work (success rate, time-to-dock, baseline/yaw consistency, and failure-modes) provide a practical foundation for statistically grounded comparison of docking supervision and controller configurations A complete hardwaresoftware stack (ROS 2 with Crazyflie/PX4 interfaces) was implemented, enabling synchronized estimation, setpoint generation, and systematic evaluation The benchmark-style metrics reported in this work (success rate, time-to-dock, baseline/yaw consistency, and failure-modes) provide a practical foundation for statistically grounded comparison of docking supervision and controller configurations A complete hardwaresoftware stack (ROS 2 with Crazyflie/PX4 interfaces) was implemented, enabling synchronized estimation, setpoint generation, and systematic evaluation The benchmark-style metrics reported in this work (success rate, time-to-dock, baseline/yaw consistency, and failure-modes) provide a practical foundation for statistically grounded comparison of docking supervision and controller configurations A complete hardwaresoftware stack (ROS 2 with Crazyflie/PX4 interfaces) was implemented, enabling synchronized estimation, setpoint generation, and systematic evaluation The benchmark-style metrics reported in this work (success rate, time-to-dock, baseline/yaw consistency, and failure-modes) provide a practical foundation for statistically grounded comparison of docking supervision and controller configurations A complete hardwaresoftware stack (ROS 2 with Crazyflie/PX4 interfaces) was implemented, enabling synchronized estimation, setpoint generation, and systematic evaluation The benchmark-style metrics reported in this work (success rate, time-to-dock, baseline/yaw consistency, and failure-modes) provide a practical foundation for statistically grounded comparison of docking supervision and controller configurations <ref:2605.

1. Robustness against transient amplification in coupled maneuvers: The progress-aware mission supervisor enforces measurable tolerance gates before advancing phases, which maintains synchronized motion to suppress transient formation during coupled leader-follower maneuvers. This allows for controlled, repeatable engagement even when velocity mismatches occur, preventing accordion transients during docking.

2. Repeatable docking under payload constraints: The platform enables a lightweight modular frame with passive magnetic latching compatible with micro-UAV payload constraints (approximately 15g). This system can execute a repeatable approach and docking behavior in real-world experiments, providing a practical testbed for modular aerial cooperation.

3. Benchmarking and comparative analysis: The system provides statistically grounded comparison of docking supervision and synchronization strategies by reporting quantitative metrics such as formation error, baseline and yaw consistency, docking success rate, time-to-dock, and failure-mode statistics. This allows researchers to systematically evaluate different control configurations.

4. Data-driven control tuning: The use of Bayesian Optimization (BO) to tune PID gains reduce overshoot and improve settling under limited thrust margins. This enables the AI system to achieve improved damping, reduced overshoot, and lower velocity at contact, which directly increases capture probability and reduces bounce-off failures.

5. Phase transition formalization: The system converts docking into a measurable, benchmarkable process with clearly defined success conditions by defining tolerances for relative position, yaw, and velocity (e.g., "eb < εb, eψ < εψ, vrel < εv). This formalizes the complex docking task into discrete stages: Approach → Align → Capture → Settle."

Reliable midair docking between small unmanned aerial vehicles (UAVs) is essential for modular aerial cooperation and manipulation, but it requires precise relative-pose control and repeatable platform under tight thrust and payload constraints. We present a dual-drone docking platform where two quadrotors operate in a leader-follower formation and dock using a lightweight modular frame with passive magnetic latching. A progress-aware mission supervisor manages phase transitions: approach, alignment, capture, and settle. This platform integrates a complete hardware-software stack (ROS 2 with Crazyflie/PX4 interfaces) and synchronized logging for benchmark evaluation. We evaluate the platform in simulation and real-world experiments using quantitative metrics such as formation error, baseline and yaw consistency, docking success rate, time-to-dock, and failure-mode statistics. The platform enables statistically grounded comparison of docking supervision and synchronization strategies and provides a practical testbed for modular aerial cooperation and repeatable midair aerial manipulation.