Hoverflie: An empirical investigation of rotor shrouds to transform micro air vehicles into multi-modal hovercraft

Technical Summary: Hoverflie: An Empirical Investigation of Rotor Shrouds to Transform Micro Air Vehicles into Multi-Modal Hovercraft

Problem Statement
Micro air vehicles (MAVs) designed for indoor or built-environment operations face severe limitations in flight duration due to low Reynolds number aerodynamics, significant induced losses, and constrained onboard energy storage. While proximity-induced aerodynamic effects (ground effect) offer a mechanism to augment lift and reduce power consumption, their application to MAV-scale multirotors remains underexplored. Existing ground effect models, often derived from full-scale helicopters or simple potential flow theories, fail to account for complex interactions in shrouded multirotor systems, such as flow recirculation, back pressure, and the "suckdown" effect (a reduction in net force at intermediate heights). Furthermore, integrating shrouds to exploit ground effect often degrades free-flight performance, creating a trade-off that prevents the development of versatile, multi-modal robots capable of both efficient surface-assisted locomotion and aerial flight.

Methodology
The authors developed "Hoverflie," a multi-modal robot based on the Crazyflie 2.1 platform, equipped with a custom, lightweight rotor-integrated shroud system. The research employed a systematic empirical approach involving:

1. Experimental Platform: A custom test rig was constructed using a 3D-printed gantry, a load cell, and a programmable interface to precisely control vehicle height and rotor duty cycle. This setup allowed for automated logging of lift forces across varying normalized heights.
2. Parametric Shroud Design: The shroud was decomposed into three functional geometric regions: the duct, the nozzle, and the intake. Each component was parameterized by diameter, height, and angular profile.
3. Fabrication: To balance rapid prototyping with weight reduction, initial geometries were 3D printed for testing, while final optimized configurations were manufactured using thin-film thermoforming (0.005" ABS) over 3D-printed molds.
4. Testing and Modeling: The team conducted parametric testing to evaluate the impact of shroud geometry on in-ground-effect (IGE) thrust, out-of-ground-effect (OGE) thrust, and transition dynamics. Experimental data was used to validate classical ground effect models (Cheeseman–Bennett) and develop a new empirical model capable of capturing the suckdown phenomenon.
5. Flight Validation: A simple mode-switching controller was implemented to demonstrate stable hovering, free-flight, and transitions between the two modes, with trajectory tracking errors quantified against a stock Crazyflie baseline.

Key Contributions
The paper presents five primary contributions:

1. Reproducible Methodology: A high-resolution experimental framework for characterizing ground effect forces in multirotor MAVs.
2. Geometric Mapping: A systematic empirical mapping between shroud geometry (duct, nozzle, intake) and aerodynamic performance, identifying optimal configurations for lift augmentation.
3. Empirical Modeling: The development of a novel empirical model that captures the suckdown effect in shrouded multirotors, addressing the limitations of classical image-theory-based models which fail to predict force dips at transition heights.
4. Multi-Modal Demonstration: The realization of a lightweight hovercraft capable of trajectory tracking and smooth hover-to-flight transitions.
5. Quantified Trade-offs: A detailed analysis of endurance trade-offs, demonstrating substantial energy savings in IGE operation against a penalty in OGE flight.

Results

• Aerodynamic Performance: An optimized shroud configuration (Duct: 47mm/15mm; Nozzle: 47mm/15mm/45°; Intake: 47mm/15mm/45°) achieved nearly three times higher lift in-ground-effect compared to the stock drone while maintaining comparable out-of-ground-effect thrust.
• Suckdown Mitigation: The study identified that nozzle angle significantly influences suckdown magnitude. While larger angles increased IGE lift, they also increased the suckdown dip during transition. The addition of an intake and the reduction of nozzle surface area helped mitigate negative entrainment effects.
• Endurance: The optimized Hoverflie exhibited a 60% increase in flight duration during in-ground-effect hovering (889s vs. 561s for the stock drone) but suffered a 30% reduction in free-flight endurance (382s vs. 544s) due to the added mass (net increase of 15.8g, ~50% of the base vehicle weight).
• Control and Tracking: The system demonstrated stable hovering and free-flight. However, tracking errors were higher during free-flight and transition maneuvers compared to the stock drone, particularly in yaw control, attributed to increased inertia and complex fluid dynamics. The transition phase showed the highest deviation (7.8 cm RMS error) due to rapid aerodynamic changes.
• Modeling: The classical Cheeseman–Bennett model failed to capture the suckdown dip. The proposed empirical model successfully fit the data, allowing for quantitative comparison of design parameters (e.g., correlating parameter B with suckdown magnitude).

Significance and Claims
The authors claim that this work provides an experimentally validated foundation for hybrid hovercraft-drone systems. By demonstrating that lightweight structural modifications can transform a standard MAV into a multi-modal vehicle, the paper establishes passive design as a practical lever for energy-aware mobility. The work highlights that while ground effect offers substantial efficiency gains for near-surface operations, it introduces specific aerodynamic challenges (suckdown) that require tailored geometric solutions rather than simple scaling of existing helicopter models. The paper concludes that such systems are feasible for applications requiring extended near-surface endurance while retaining the capability for full aerial mobility, provided that control strategies account for the distinct aerodynamic regimes and transition dynamics.