Robotic Manipulation in Space

This thesis addresses several challenges related with robotic manipulation in space, specifically focusing on a satellite equipped with a manipulator arm. It presents several novel designs to overcome the limitations of existing state-of-the-art solutions in the topics of robust control, online trajectory generation and cooperative manipulation. First, an in-depth discussion is provided on the kinematics and dynamics of space manipulator systems and manipulated target satellites, along with an overview of the actuation system, disturbances and uncertainties affecting the motion of both the satellite and the target. Next, two innovative controllers, one based on sliding mode control and the other on indirect adaptive control, are developed to perform trajectory tracking under disturbances and model uncertainties. The sliding mode control approach focuses on enhancing controller performance in the presence of large attitude displacements of the base satellite. The indirect adaptive control solution mitigates the effects of uncertainties, actuator saturation, and loss of actuator effectiveness. The proposed adaptive law preserves the physical consistency of the estimated inertial parameters, which enables the use of an indirect approach. Then, a simple strategy for generating test trajectories, using inverse kinematics, and a novel online motion planning algorithm, utilizing a nonlinear model predictive control framework, are introduced. In the online solution, special attention is given to maintaining real-time performance and ensureing collision avoidance between the manipulator and the target. The next chapter presents a distributed adaptive control strategy for multiple robotic agents cooperatively manipulating a payload in space. An indirect adaptive control scheme is proposed to handle uncertainties in the manipulated object and limited actuation capabilities, featuring an anti-windup mechanism for the estimated inertial parameters in the case of actuator saturation. This approach also ensures physical consistency of the estimated parameters. Additionally, a dynamic input allocation strategy is introduced to effectively distribute the control effort among the agents. With the exception of the online trajectory generation problem, the stability of the closed-loop systems is proven theoretically. Additionally, the performance and robustness of all proposed solutions are evaluated through realistic simulations based on multibody models built in Matlab-Simulink using the Mechanics toolbox. The performance and robustness of the online trajectory generation algorithm is compared with that of an offline solver, while the other algorithms are benchmarked against existing state-of-the-art solutions. The thesis concludes with final remarks that summarize the key advantages and limitations of the proposed solutions.