Improved crane robot control for confined space in-wing teleoperated inspection

Abstract

Visual inspection of confined spaces such as aircraft wings is ergonomically challenging for human mechanics. This dissertation presents a novel crane robot that can travel the entire span of the aircraft wing, enabling mechanics to perform inspection from outside of confined spaces. However, teleoperation of the crane robot can still be a challenge due to the need to avoid obstacles in the workspace and potential oscillations of the camera payload. Therefore, a teleoperation assistance is developed using the differential flatness of the crane-robot dynamics for designing reduced oscillation, collision-free time trajectories of the camera payload for use in teleoperation. Autonomous experiments verify the efficacy of removing undesired oscillations while teleoperation experiments demonstrate that the controller eliminated collisions when 12 participants performed an inspection task with the use of proposed trajectory selection when compared to the case without it. Moreover, even discounting the failures due to collisions, the proposed approach improved task efficiency. Extended to a crane-robot-based active vision system for in-wing confined space inspection, the crane robot enables teleoperated manipulators to navigate around in-wing systems while providing a dynamic view of potential collisions to the teleoperator. Sampling-based techniques, such as the model predictive path integral control (MPPI), are well suited to optimize the input to the active vision system for precisely tracking the teleoperated robot while maintaining constraints and avoiding obstacles. However, correctly selecting the input sample distribution for MPPI is often difficult for general systems, which can be addressed by iteratively optimizing over long preview-time horizons, increasing computational load and inducing tracking delays. Therefore, an output-sampled MPPI (oMPPI) is developed using inversion-based control, where the input to the approach becomes the desired output (reference position) for general invertible nonlinear systems, which enables sampling of the system output rather than the input. An advantage of the proposed oMPPI is that the selection of the reference output's sample distribution can be based on the desired output of the system, such as the trajectory of the teleoperated robot the active-vision case. Additionally, optimality with the proposed oMPPI is proved and conditions are provided for feasibility of the inverse. Experimental results show that oMPPI enhances sampling efficiency over MPPI by reducing the required preview time. Furthermore, oMPPI enables precision tracking by reducing both oscillations and tracking errors during teleoperation with the active vision system.