Presented By: Michigan Robotics
Leveraging Intentional Collisions to Enhance Aerial Manipulation
Mark Nail, PhD Defense
Committee Chairs: Brent Gillespie and Ella Atkins
Zoom Link: https://umich.zoom.us/j/95579522863
Abstract:
Uncrewed Aircraft Systems (UAS) have become essential in tasks that are too difficult or dangerous for humans. This dissertation was initially inspired by the challenge of making rainforest tree sampling safer for botanists, who often need to climb dangerously tall trees. Building on this motivation, we developed a new approach for aerial manipulation that enables interaction with objects of unknown dynamics through intentional collisions using a pogo-stick payload. The key insight is that intentional collisions consolidate all potential disturbance forces into a single impulse, the outcome of which is hopefully simpler to predict. Given assumptions on the dynamics of the interaction object, the map from pre- to post-impact states (collision response) is known, providing an opportunity to choose pre-impact states that will result in recoverable post-impact states.
The use of intentional collisions with a UAS presents unique challenges due to underactuation. Collision outcomes are set by strike velocity and orientation, yet hitting a desired point requires full position control. Traditional UAS actuator configurations cannot provide both full position and attitude control authority. This dissertation presents several approaches to mitigate underactuation and enable aerial manipulation with intentional collisions.
To begin, this dissertation develops the strategy of using intentional collisions in a flight mode called Velocity Matching. The Velocity Matching flight mode starts in two phases: launch phase and then the Velocity Matching phase. During the launch phase, the UAS is sped up horizontally to "launch" the vehicle, assumed thereafter to behave ballistically, towards the target. After launch, the Velocity Matching phase is entered, and thrust is limited to the bare minimum thrust required for attitude control. During the Velocity Matching phase the minimal thrust attitude control is used to align the center-of-mass velocity vector of the UAS with the pogo-stick’s orientation at strike. Using simulation and flight tests, we generate a map from pre-launch modes states into recoverable post-impact states, allowing for planned strikes.
Second, we refine the flight control from Velocity Matching to a cascaded position–attitude switching controller (called Switching Control). In the Switching Control mode, the UAS flies in position cascaded PID control for most of the trajectory, then switches at the last moment to attitude cascaded PID control for pre-strike alignment. We also provide a tool to predict the best switching time from an initial displacement, and we validate the approach experimentally.
Finally, we develop a new control approach involving an actuated pogo-stick payload controlled via a partial feedback linearizing controller to regulate both UAS position and pogo-stick orientation. After validating the actuated pogo-stick in simulation, we built a physical cable-driven version that slides along a hemispherical base. This mechanism provides two-degree-of-freedom orientation control (pitch and roll compensation) and keeps the pogo-stick perpendicular to the UAS center of mass, enabling more favorable, recoverable collision responses.
Zoom Link: https://umich.zoom.us/j/95579522863
Abstract:
Uncrewed Aircraft Systems (UAS) have become essential in tasks that are too difficult or dangerous for humans. This dissertation was initially inspired by the challenge of making rainforest tree sampling safer for botanists, who often need to climb dangerously tall trees. Building on this motivation, we developed a new approach for aerial manipulation that enables interaction with objects of unknown dynamics through intentional collisions using a pogo-stick payload. The key insight is that intentional collisions consolidate all potential disturbance forces into a single impulse, the outcome of which is hopefully simpler to predict. Given assumptions on the dynamics of the interaction object, the map from pre- to post-impact states (collision response) is known, providing an opportunity to choose pre-impact states that will result in recoverable post-impact states.
The use of intentional collisions with a UAS presents unique challenges due to underactuation. Collision outcomes are set by strike velocity and orientation, yet hitting a desired point requires full position control. Traditional UAS actuator configurations cannot provide both full position and attitude control authority. This dissertation presents several approaches to mitigate underactuation and enable aerial manipulation with intentional collisions.
To begin, this dissertation develops the strategy of using intentional collisions in a flight mode called Velocity Matching. The Velocity Matching flight mode starts in two phases: launch phase and then the Velocity Matching phase. During the launch phase, the UAS is sped up horizontally to "launch" the vehicle, assumed thereafter to behave ballistically, towards the target. After launch, the Velocity Matching phase is entered, and thrust is limited to the bare minimum thrust required for attitude control. During the Velocity Matching phase the minimal thrust attitude control is used to align the center-of-mass velocity vector of the UAS with the pogo-stick’s orientation at strike. Using simulation and flight tests, we generate a map from pre-launch modes states into recoverable post-impact states, allowing for planned strikes.
Second, we refine the flight control from Velocity Matching to a cascaded position–attitude switching controller (called Switching Control). In the Switching Control mode, the UAS flies in position cascaded PID control for most of the trajectory, then switches at the last moment to attitude cascaded PID control for pre-strike alignment. We also provide a tool to predict the best switching time from an initial displacement, and we validate the approach experimentally.
Finally, we develop a new control approach involving an actuated pogo-stick payload controlled via a partial feedback linearizing controller to regulate both UAS position and pogo-stick orientation. After validating the actuated pogo-stick in simulation, we built a physical cable-driven version that slides along a hemispherical base. This mechanism provides two-degree-of-freedom orientation control (pitch and roll compensation) and keeps the pogo-stick perpendicular to the UAS center of mass, enabling more favorable, recoverable collision responses.
Explore Similar Events
-
Loading Similar Events...
