Abstract:
Uncrewed Aerial Vehicles (UAVs) have attracted the interest of researchers, industry, and the general public in many applications. Noticing that high-altitude tasks sometimes require active interaction with the environment, there have been more and more works focusing on aerial manipulation recently. Each of them has demonstrated the ability to use a specific aerial manipulator platform to achieve one specific complex aerial manipulation task. These studies demonstrate impressive capabilities but lack versatility across multiple aerial manipulation tasks.

In this thesis, we aim to answer the question: how to achieve universal aerial manipulation? Specifically, how to achieve multiple different aerial manipulation tasks in different scenarios, with a general aerial manipulator hardware platform and a general aerial manipulation algorithm pipeline?

We begin by developing the hardware platform. Specifically, we pioneered the use of vision-based tactile sensors on a fully actuated hexarotor and demonstrated the whole system on a specific aerial interaction task. Next, we targeted a more general aerial interaction task: simultaneously tracking time-varying contact force in the surface normal direction and motion trajectories on tangential surfaces. We propose a pipeline that includes a contact-aware trajectory planner to generate dynamically feasible trajectories, and a hybrid motion-force controller to track such trajectories.

Finally, we propose a comprehensive framework designed to address a broader range of aerial manipulation tasks. This framework introduces a versatile aerial manipulator platform capable of achieving different tasks, including pick and place, insertion, and making point contact precisely. Central to our approach is a hierarchical algorithm pipeline composed of decision-making, planning, and control modules. Having demonstrated the effectiveness of the planning and control module, we will mainly focus on the decision-making module, where we propose a continuous research direction: from shared autonomy to full autonomy. We first propose a pipeline to adopt human operator command as the task planner for the decision-making module. Noticing that directly using the user’s command will lead to a suboptimal trajectory, we propose to develop a human command prediction model. We fuse the prediction command and the user’s raw command for a more robust task planner. Based on that, we propose to achieve a full-autonomous system by further developing the human operation prediction mode so that the fusion model can be replaced by only the prediction model.

We believe the proposed system, including both the hardware system and algorithm pipeline, will enable the aerial manipulators the ability to achieve multi-task in different scenarios, leading it toward achieving universal aerial manipulation.

Thesis Committee Members:
Guanya Shi, Co-Chair
Sebastian Scherer, Co-Chair
Oliver Kroemer
Giuseppe Loianno, New York University