TL;DR
PerchMobi$^3$ is a versatile multi-modal robot capable of flying, driving, and climbing vertical walls by utilizing a clever "power-reuse" mechanism. By repurposing its four ducted fans to provide both aerial thrust and negative-pressure wall adhesion, the design eliminates the need for heavy, dedicated suction pumps, resulting in a lightweight, high-payload platform for complex environments.
What Problem Does This Paper Address?
Creating a robot that can seamlessly transition between flying, driving, and wall-climbing is a "holy grail" for inspection and search-and-rescue. However, most existing designs suffer from a "complexity tax." To stick to a wall, robots typically need dedicated hardware—like vacuum pumps or adhesive grippers—which adds significant weight and reduces battery life. This extra mass makes the robot a poor flyer, while the complexity makes the control systems brittle.
The authors of PerchMobi$^3$ set out to break this trade-off by asking: Can we make the propulsion system do double duty?
Methods
The researchers developed a quad-fan, four-wheeled robot built around an integrated carbon-fiber frame. The core innovation is the propulsion-adhesion power-reuse mechanism:
- Hardware Design: The robot uses four ducted fans mounted on a cross-shaped frame. The underside of the frame (Surface B) features a sealing sponge. When the robot is against a wall, these fans spin to create a low-pressure zone inside the frame, "sucking" the robot onto the surface.
- Multi-Modal Control: The team established a unified modeling and control framework for four distinct states:
- Ground Mode: Differential driving using four active wheels.
- Wall-Climbing Mode: Using the fans for adhesion while the wheels provide vertical and lateral locomotion.
- Flight Mode: Standard quadrotor-style flight.
- Transition Mode: A specialized "airframe flipping" maneuver that uses asymmetric thrust to lift the robot from the ground onto a vertical wall.
- Electrical Architecture: Powered by either a tether or a 6S LiPo battery, the system uses an STM32F427 controller and an IMU to manage high-frequency (400Hz) orientation updates and motor control.
Datasets
As this is a robotics hardware and control paper rather than a pure machine learning study, the authors did not use standard training datasets. Instead, they generated original experimental data to validate their models:
- Adhesion Characterization: A dataset mapping throttle input to comprehensive adhesion capacity (N), resulting in a quadratic fitting curve used for precise control.
- Trajectory Data: 3D, X-Y, Y-Z, and X-Z trajectory points were recorded during experimental trials to prove the stability of cross-mode transitions.
Key Insights and Ideas
- The "Aha" Moment of Power Reuse: Ducted fans are inherently efficient at moving large volumes of air. By placing them within a sealed cavity, they become highly effective vacuum generators. This dual-use eliminates the weight of a separate vacuum motor, which is usually the heaviest component of a wall-climber.
- Asymmetric Torque for Flipping: To transition from the ground to a wall, the robot doesn't just drive forward. It uses the fans furthest from the wall to generate a pitching moment, "flipping" the front wheels up until they make contact with the vertical surface.
- Elastic Buffering: The tires are designed with a 0.5mm clearance from the frame. This provides a "suspension" effect during the violent transition from flight/ground to wall, protecting the carbon fiber frame from impact damage.
Key Results
- Weight & Payload: The robot weighs 3.2kg and successfully demonstrated stable wall-climbing while carrying a payload of over 2kg—a significant feat for a multi-modal platform.
- Adhesion Strength: At 100% throttle, the system generates an effective adhesion force of 165.68N, providing a massive safety margin for vertical operations.
- Versatility: Experiments confirmed the robot could traverse narrow gaps (ground mode), leap over obstacles (flight mode), and climb vertical facades (wall mode) with smooth, fan-assisted transitions.
Limitations and Future Work
While PerchMobi$^3$ is a major step forward, the authors acknowledge several areas for improvement:
- Autonomy: The current prototype is operated via remote control. Future versions will require onboard perception (LiDAR/Cameras) and path planning to navigate autonomously.
- Energy Efficiency: While power-reuse is efficient, the team plans to further optimize the coordination between fans and wheels to extend battery life.
- Surface Adaptability: The current sealing mechanism works best on relatively smooth surfaces. Testing on highly unstructured or porous surfaces (like brick or cracked concrete) remains a challenge for future iterations.
- Tethering: Many tests were performed with a tether; moving to fully untethered, long-endurance missions is the next logical milestone.
Summary updated