Development of a Versatile Modular Platform for Aerial Manipulators

Nikolaos Evangeliou, Αθανάσιος Τσουκαλάς, Nikolaos Giakoumidis, Steffen Holter, Anthony TzesIntechOpen 2020 · DOI · OpenAlex

Abstract: The scope of this chapter is the development of an aerial manipulator platform using an octarotor drone with an attached manipulator. An on-board spherical camera provides visual information for the drone’s surroundings, while a Pan-Tilt-Zoom camera system is used to track targets. A powerful computer with a GPU offers significant on-board computational power for the visual servoing of the aerial manipulator system. This vision system, along with the Inertial Management Unit based controller provides exemplary guidance in confined and outdoor spaces. Coupled with the manipulator’s force sensing capabilities the system can interact with the environment. This aerial manipulation system is modular as far as attaching various payloads depending on the application (i.e., environmental sensing, facade cleaning and others, aerial netting for evader-drone geofencing, and others). Experimental studies using a motion capture system are offered to validate the system’s efficiency.

Introduction

While unmanned aerial vehicles (UAVs) have become ubiquitous in surveillance and cinematography, transitioning from passive observation to active physical interaction—aerial manipulation—remains a significant technical challenge. Most commercial drones lack the payload capacity and computational power required for real-time visual servoing and the management of contact forces. Furthermore, the forces exerted by a robotic arm can easily destabilize a standard flight controller. This research presents a heavy-lift, modular octarotor platform designed to overcome these hurdles through high-thrust hardware, redundant localization, and sophisticated computer vision.

Platform Design and Methodology

Mechatronic Architecture

The foundation of the platform is a custom-built H-frame octarotor. To achieve the necessary lift for heavy manipulators, the design employs a coaxial motor configuration, providing a massive 40 kg maximum take-off weight and approximately 30 minutes of flight time.

Structural design of the H-frame octarotor platform.

Modularity and Manipulation

Central to the platform's versatility is its modular mounting system. Using quick-release clamps and specialized mounts, the drone can be rapidly reconfigured for diverse tasks. For manipulation, the system integrates a Kinova Gen 2 Assistive 7-Degree-of-Freedom (DoF) robotic arm. This arm is equipped with torque sensors at every joint, allowing the system to "feel" and respond to environmental contact forces during complex tasks like ultrasound scanning.

Exploded view of the modular mounting system and quick-release clamps.

Dual-Vision and Localization

The platform utilizes a unique "detect-then-zoom" vision strategy. A 360° spherical camera provides constant situational awareness, while a Pan-Tilt-Zoom (PTZ) camera tracks specific targets with high resolution. On-board GPU-accelerated algorithms process these feeds for visual object tracking (VOT), isolating moving targets through homography-based background subtraction.

For localization, the system employs a multi-modal approach:

  • Outdoors: Real-Time Kinematic (RTK) GNSS provides centimeter-level horizontal precision.
  • Indoors: A Motion Capture System (MoCaS) utilizing 24 Vicon cameras injects high-frequency (25 Hz) positioning data into the flight stack to compensate for the lack of GPS.

Key Insights and Results

Localization Performance

Experimental results demonstrate that RTK injection significantly enhances stability, reducing the standard deviation in horizontal positioning from approximately 0.74m to just 0.05m. However, the study notes that RTK offers negligible improvements in vertical (Z-axis) accuracy, necessitating the continued use of barometers or laser altimeters for altitude hold.

Collaborative Accuracy

In multi-drone scenarios, the system achieved a relative 3D localization accuracy of 2.2 cm using marker-based vision. This precision is vital for collaborative manipulation tasks where multiple platforms must coordinate their movements in close proximity.

Physical Interaction

The platform successfully demonstrated contact-based ultrasound scanning on surfaces inclined at 45°. This task validated the system's ability to maintain a stable hover while the 7-DoF manipulator exerted controlled forces (between 10 and 100 N) against a target.

Conclusion and Future Work

The developed octarotor provides a robust foundation for complex aerial manipulation. While the current system handles significant contact forces, future work will focus on implementing advanced control techniques, such as Model Predictive Control (MPC) and Adaptive Backstepping, to better manage the dynamic coupling between the manipulator and the drone frame. Additionally, the researchers aim to increase motor control rates to $\ge$ 1 kHz to improve responsiveness during high-precision contact operations in confined environments.

Summary updated

Abstract

The scope of this chapter is the development of an aerial manipulator platform using an octarotor drone with an attached manipulator. An on-board spherical camera provides visual information for the drone’s surroundings, while a Pan-Tilt-Zoom camera system is used to track targets. A powerful computer with a GPU offers significant on-board computational power for the visual servoing of the aerial manipulator system. This vision system, along with the Inertial Management Unit based controller provides exemplary guidance in confined and outdoor spaces. Coupled with the manipulator’s force sensing capabilities the system can interact with the environment. This aerial manipulation system is modular as far as attaching various payloads depending on the application (i.e., environmental sensing, facade cleaning and others, aerial netting for evader-drone geofencing, and others). Experimental studies using a motion capture system are offered to validate the system’s efficiency.
BibTeX
@inproceedings{evangeliou2020development,
  author = {Nikolaos Evangeliou and Αθανάσιος Τσουκαλάς and Nikolaos Giakoumidis and Steffen Holter and Anthony Tzes},
  title = {{Development of a Versatile Modular Platform for Aerial Manipulators}},
  booktitle = {IntechOpen},
  year = {2020},
  doi = {10.5772/intechopen.94027},
}
Figure