Research

Aerial Grasping and Object Interaction

My current work at Yale focuses on grasping and manipulation of objects by robot helicopters. We want to extend the capabilities of hovering vehicles to include direct physical interaction with objects and the environment. This task is challenging because of the inherently limited positioning performance of unstable platforms such as helicopters.



Our solution is to use an adaptive compliant manipulator. This class of robot hand is well-suited to the aerial manipulation task. Unlike a rigid gripper, it allows for large positioning errors during grasping. It also permits motion of the object relative to the body of the robot, improving the stability of the system. This lightweight gripper module was specially designed to fit between our helicopter's skids.



The coupled dynamics of the helicopter-object system are critical. When a hovering robot grasps an object but has not yet lifted it, it becomes mechanically linked to ground. The stiffness of the linkage determines whether this system is stable - if it is unstable, the robot will oscillate and eventually crash. By analysing these coupled flight dynamics, I have identified regions of stability for aircraft control and stiffness parameters.



By correctly designing the end-effector, it is possible to build aerial manipulation systems that will be inherently stable during object contact under the same attitude controller used in free flight. This is important, as the uncertainty in helicopter position makes switching between different controllers for specific contact regimes dicey.



Last year we demonstrated the first successful in-flight grasping and retrieval of an unstructured object by an unmanned helicopter, and since then we have picked up and manipulated a variety of objects in different grasps and contact configurations.



In addition to coupled mechanical stability, there are many other essential aspects: static flight stability, ground effect and ground-vortex rollup during capture, dynamic stability with changing load, grasp stability, and more. I will add more to this page as I have time. Until then, please enjoy these pictures of the Yale Aerial Manipulator at work:

Quadrotor Dynamics and Control

My doctoral thesis project focussed on the modelling, dynamics and control of a large quadrotor. We determined to build a large vehicle with the expectation of mounting a 1kg SICK laser scanner and vision sensors on board, but while retaining a compact 1m envelope. This was a very challenging goal, as the energetic requirements for a vehicle that weight in such a small space pushed the limits of the battery, motor and rotor technology.



A significant component of my research was the design of flexible, thin rotor blades that twist under aerodynamic load. I developed a distortion model and simulator to compute the degree of twist induced in the rotors. The rotors are designed with an initial pre-twist such that they twist into the ideal angle of attack at steady-state. This makes the rotors far more efficient than would be possible with uncompensated blades or sufficiently thick airfoil sections.





There is much much more to come! Updates as I scrape together the time to fill out this page:
* Drive system design
* Dynamic Modelling
* Control Design - Design for Control
* Non-linear Control Estimator
* Avionics and ground station software
* Flight Tests

Back

Alternate contact: paul.pounds_at_gmail.com

Updated 2010-07-07