This paper presents the design and experimental assessment of the control system for the UX-1 robot, a novel spherical underwater vehicle for flooded mine tunnel exploration. Propulsion and maneuvering are based on an innovative manifold system. First, the overall design concepts of the robot are presented. Then, a theoretical six degree-of-freedom (DOF) dynamic model of the system is derived. Based on the dynamic model, two control systems have been developed and tested, one based on the principle of nonlinear state feedback linearization and another based on a finite horizon linear quadratic regulator (LQR). A series of experimental tests have been carried out in a controlled environment to experimentally identify the complex parameters of the dynamic model. Furthermore, the two proposed controllers have been tested in underwater path tracking experiments designed to simulate navigation in mine tunnel environments. The experimental results demonstrated the effectiveness of both the proposed controllers and showed that the state feedback linearization controller outperforms the finite horizon LQR controller in terms of robustness and response time, while the LQR appears to be superior in terms of fall time.

No abstract

Recurrent Neural Networks (RNNs) are increasingly being used for model identification, forecasting and control. When identifying physical models with unknown mathematical knowledge of the system, Nonlinear AutoRegressive models with eXogenous inputs (NARX) or Nonlinear AutoRegressive Moving-Average models with eXogenous inputs (NARMAX) methods are typically used. In the context of data-driven control, machine learning algorithms are proven to have comparable performances to advanced control techniques, but lack the properties of the traditional stability theory. This paper illustrates a method to prove a posteriori the stability of a generic neural network, showing its application to the state-of-the-art RNN architecture. The presented method relies on identifying the poles associated with the network designed starting from the input/output data. Providing a framework to guarantee the stability of any neural network architecture combined with the generalisability properties and applicability to different fields can significantly broaden their use in dynamic systems modelling and control.

Autonomous underwater gliders (AUGs) are currently deployed in oceans throughout the globe and are recording real-time, in-situ data. Simulating AUGs is rendered particularly difficult by the identification of the underlying dynamic model, as these vehicles embed several internal movable components. Acausal simulators can significantly improve the possibility to study and understand the dynamics of this class of vehicles and can in turn support the design of more robust control systems. In this paper, an open-source simulator architecture designed in OpenModelica is proposed to simulate underwater gliders. The validation is carried out using two different AUGs models, a ROGUE and a Seawing. The vehicle dynamics is firstly compared with analytical results and, following, with values obtained by means of another simulator. Further steps will entail comparing the dynamics of a simulated Seaglider with real deployment data publicly available. In this work, some of the main hydrodynamic and geometrical properties of a Seaglider are identified, computed through Computational Fluid Dynamics (CFD) analyses and retrieved from the mission ballast sheet. Overall, the developed model is expected to enhance gliders' control strategies, thus improving their performance and mitigating incidents such as being carried away by undesired ocean currents.

No abstract