Target Trajectory Prediction-based Object Handover in a 3D Heterogeneous Robot System
von Wei LuoThis work is devoted to contributing multiple key-functionalities for cooperation within a heterogeneous robot group. Specifically, a hypothetical scenario is investigated, where an aerial manipulator should hand over an object from a moving mobile robot on the ground. In this cooperation, both robots are not centrally controlled. In contrast, the mobile robot is controlled separately, and the aerial manipulator can only observe the mobile robot's previous movements, and it must collaborate with the mobile robot in a more energy-efficient manner.
The first contribution of this thesis is to provide solutions for predicting the future trajectory of the observed mobile robot based on its previous movements. Once the likely future trajectory of the mobile robot is acquired, the next challenge is to plan a trajectory of the aerial manipulator to approach and implement the desired handover cooperation in a time-optimal manner without validating the aerial manipulator's dynamics. In the proposed DMCC framework, the system dynamics is constrained with the discrete variational Lagrangian mechanics, which yields reliable estimation results. Moreover, the handover opportunities are determined and arranged automatically based on the desired complementarity constraints. The last contribution of this thesis is dedicated to controlling the aerial manipulator based on nonlinear model predictive control combined with an augmented dynamic model employing Gaussian processes as a nonparametric regression model, which allows the proposed NMPC control framework to determine the optimal control inputs for the aerial manipulator and achieve a stable flight performance.
The first contribution of this thesis is to provide solutions for predicting the future trajectory of the observed mobile robot based on its previous movements. Once the likely future trajectory of the mobile robot is acquired, the next challenge is to plan a trajectory of the aerial manipulator to approach and implement the desired handover cooperation in a time-optimal manner without validating the aerial manipulator's dynamics. In the proposed DMCC framework, the system dynamics is constrained with the discrete variational Lagrangian mechanics, which yields reliable estimation results. Moreover, the handover opportunities are determined and arranged automatically based on the desired complementarity constraints. The last contribution of this thesis is dedicated to controlling the aerial manipulator based on nonlinear model predictive control combined with an augmented dynamic model employing Gaussian processes as a nonparametric regression model, which allows the proposed NMPC control framework to determine the optimal control inputs for the aerial manipulator and achieve a stable flight performance.