Design, modeling, and predictive control of aerial physical interaction towards proactive maintenance

• Main Author: Kocer, Basaran Bahadir
• Other Authors: Seet Gim Lee, Gerald
• Format: Theses and Dissertations
• Language: English
• Published: 2019
• Subjects: Engineering::Mechanical engineering
• Online Access: https://hdl.handle.net/10356/85159; http://hdl.handle.net/10220/50267
• Institution: Nanyang Technological University

Modern cities are dependent on basic services like water and sewer systems including tunnels, which are supported by a pervasive infrastructure. The structural condition of these places is prone to deterioration. To protect the integrity of such systems, regular inspection is required to detect the onset of damage and material failure. It is envisaged that several robotic systems may provide a solution for the inspection task. When the inspection is purely visual, cameras might provide the necessary functionality. However, the subsurface inspection may require a different approach. It is explored that unmanned aerial vehicle (UAV) with contacting probes may be an option. Deploying a UAV to perform contact inspection can be a more demanding task. First, the inspection tool shifts the center of mass and moment of inertias. Second, the terminal velocity needs to be regulated for soft contact. This phase necessitates the system to be working in faster cycles as compared to the free-flight regime. The last challenge, posed by the interaction tool, comes from the sliding phase on the surface because bouncing occurs while it moves. In addition, ceiling effect where variable rotor wake degrades the performance of tracking predefined trajectory, and an unknown surface that must be included in the control algorithm and interaction tool design. The available approaches can handle the UAV control while it flies in free-flight. However, the challenges associated with the interaction require the system to be more responsive, adaptive and resilient. Since the level of interaction requires a force bound, the system has to explicitly consider this limit. The system needs to maintain contact in a certain force range to collect the data while not crashing the sensor at the top. The current state of the art considering constraints makes use of two individual models interaction problems; free-flight and contact phase models that bring additional complexity. Moreover, nominal optimization-based approaches are considered in the UAV control for the physical interaction tasks, wherein the system is lacking the ability to take external forces, changing parameters and unmodelled dynamics into account. In this thesis, a predictive UAV control is proposed to investigate interactions during the contact phase operation in close proximities to the surroundings. The main contribution of this study lies in two aspects: modeling and control of the UAV interaction problem. As opposed to multi-model approaches, a modeling of the whole system is presented with a centralized algorithm in which free flight, target approach, interaction, and sliding have been included. Additionally, a constraint optimization-based algorithm is designed to identify the external forces coming from the interaction tool and environment. It has shown through experiments that the proposed approach is efficient in terms of the optimized performance and applicability. The external force information increases the model size but the efficient open-source solver is adapted to leverage the problem, which consists of a nonlinear predictive controller and nonlinear moving horizon estimator, in milliseconds. In the proposed approach; the disturbances, changing parameters and unmodelled dynamics can be represented through external forces. Therefore; it does not require a precise model, external force measurement, and exact information about the environment. The hypothesized force-based centralized approach implements an optimization method by explicitly considering the external forces in the algorithm to solve the contact-based nondestructive inspection operation in the air. The results have shown that the disturbance and offset caused by the ceiling and tool effects are eliminated on the vertical axis and mitigated on the planar axes. Since there is a direct actuation on the vertical channel, it is achieved to suppress the external forces (in overall 70%). For the planar axes, the actuation is driven by the attitude angles which resulted in an attenuation of the external forces (up to 45%). It is illustrated that the selected configuration in which an elastic compliance mechanism is attached to the UAV is a pioneer to inspect surrounding environments.