Aerial Manipulation for Pick-and-Place Applications
Author
Skjelsager, Rasmus Nymark
Term
4. term
Education
Publication year
2023
Abstract
This thesis investigates the design, modeling, and state estimation of an unmanned aerial manipulator for pick-and-place tasks. It addresses the question: how can an aerial manipulator be designed, modeled, and controlled to accomplish pick-and-place? A prototype was developed comprising a quadrotor base and a lightweight two-link manipulator; mechanical parts were designed for low mass and stiffness (using FEA and topology optimization), and electronics and sensors (IMU, encoders, camera) were integrated. The kinematics and dynamics of the quadrotor, manipulator, and combined system were derived using standard robotics methods (Euler–Lagrange and Newton–Euler), alongside sensor models and estimation of key physical parameters. A 6-DOF camera system was implemented and used within an Unscented Kalman Filter–based state estimator with quaternion representation. Due to time constraints, no controller was implemented. Simulations and preliminary tests indicate that the Euler–Lagrange dynamic model matches system behavior well, whereas corrupted camera measurements degraded the estimator and prevented reliable state estimation. The thesis thus provides a validated model, parameter estimates, and an estimation framework as a step toward closed-loop aerial manipulation.
Denne afhandling undersøger design, modellering og tilstandsestimering af en ubemandet aerial manipulator til pick-and-place-opgaver. Arbejdet tager udgangspunkt i spørgsmålet: hvordan kan en aerial manipulator designes, modelleres og styres, så den kan udføre pick-and-place? Der er udviklet en prototype bestående af en quadrotor som base og en let to-leddet manipulator; de mekaniske dele blev designet med fokus på lav masse og stivhed (bl.a. ved hjælp af FEA og topologioptimering), og elektronik samt sensorer (IMU, encodere og kamera) blev integreret. Kinematik og dynamik for quadrotor, manipulator og det samlede system er beskrevet ved hjælp af standardmetoder fra robotik (Euler–Lagrange og Newton–Euler), suppleret med sensormodeller og parameterestimering af centrale fysiske størrelser. Et 6-DOF kamerasystem blev implementeret og anvendt i en tilstandsestimator baseret på et Unscented Kalman Filter med kvaternionrepræsentation. Af tidsmæssige grunde blev der ikke implementeret en regulator. Simulationer og indledende tests indikerer, at Euler–Lagrange-modellen beskriver systemets dynamik godt, mens korrupte kameramålinger nedbryder estimatorens ydeevne og forhindrer pålidelig tilstandsestimering. Afhandlingen leverer dermed en valideret model, parameterestimater og et estimeringsgrundlag som næste skridt mod lukketkreds-styring af aerial manipulation.
[This apstract has been generated with the help of AI directly from the project full text]