Abstract:
This dissertation investigates the environmental perception of autonomous underwater robots. Autonomous underwater robots have received increasing attention among scientific and engineering societies for their superior performances and promising future in marine environments. A key challenge to marine autonomy is the sensing of the background flow environment, which is important for navigation and motion control of a robot. This dissertation proposes a dynamic mode decomposition (DMD)-based flow sensing method inspired by the fish's lateral line system. DMD describes the dynamic flow field using the combination of a series of DMD modes that represent the core features of a certain flow field. This new reduced-order flow model not only captures the dynamics of the flow field, but also decreases the calculation burden; thus providing a control-oriented way to model the flow field. This flow model is integrated with a Bayesian filter to develop a DMD-based dynamic flow sensing method, which is potentially applicable to many and various complicated dynamic flow fields for arbitrarily shaped underwater robots. The number and location of sensors will both affect the flow sensing performance, because different sensor numbers and different positions contain different volumes of flow field information. The effects of these two designs in the proposed flow sensing algorithm will be investigated. Due to the harsh and unknown aquatic environment, the question of how autonomous underwater robots should navigate and maneuver, especially in a flow patterns changing environment, is still largely open. This dissertation presents a systematic background flow sensing framework to estimate surrounding flow fields with changing flow patterns and without considering transient effects. The proposed method first utilizes distributed pressure measurements of robots to determine the flow pattern/model around robots based on fast Fourier transform (FFT) spectrum analysis and then uses recursive Bayesian estimation and dynamic mode decomposition (DMD)-based modeling to identify model parameters. This algorithm seeks to detect a flow change around robots in flow pattern changing environments, which will greatly expand the application scope of the existing flow sensing methods to real-world scenarios.
