The lack of GPS signal in the underwater environment poses limitations in terms of localization and navigation of mobile robots. Strategies based on acoustic localization systems are employed to improve underwater navigation. In this paper we describe a first step towards the development of a marine system of systems involving autonomous mobile nodes. The approach relies on communication networking between an Autonomous Surface Vehicle (ASV), equipped with an Ultra Short BaseLine (USBL) device, and an Autonomous Underwater Vehicle (AUV). An active acoustic communication protocol prioritizes the AUV positioning rate, whereas a model-based navigation filter handles the delayed measurements caused by the acoustic communication latency. The system has been tested in a real marine environment to analyze its behavior and the quality of the navigation estimation. The experimental results show that the navigation algorithm on-board the AUV provides an estimate of its position with an error of a few meters with respect to the GPS ground-truth, over a total path of approximately 210m, exploiting acoustic positioning data provided by the ASV in order to limit drift problems.

Coordinated control of marine vehicles poses challenging problems, among them the possibility to use the vehicles, in addition to their nominal mission, also to achieve a relative localization task. To the purpose, the use of a surface, GPS-equipped, vehicle and one or several underwater vehicles may be envisaged. The latters can communicate among them by acoustic modems; those devices can also be used as ranging measurement units thus providing an additional information that, together with the common sensor equipment for marine vehicles, might be used for relative localization. This paper investigates how the vehicles' movements can be commanded in order to help the relative localization by a proper analysis of the system observability and a corresponding proper definition of the vehicles movements; critical situations for the relative observability, that corresponds to common movements, are then avoided. Numerical simulations on the mathematical model of the Fòlaga hybrid underwater vehicle confirm the effectiveness of the proposed coordinated behavioral approach.

This paper introduces a control-oriented model for a class of autonomous underwater vehicles (AUVs) characterized by an actuation system that includes propellers, internal weight motion and ballast charge/discharge. Since this kind of actuation combines the actuation of both oceanic gliders and self-propelled AUVs, these vehicles are termed hybrid vehicles. The modelling results are employed in the definition of an adaptive backstepping control, using a fuzzy approximator in terms of Gaussian function in order to cope with model uncertainties and disturbances. Lyapunov stability of the control is proved and simulations are shown for a typical water sampling operation.

Monitoring ocean dynamics is extremely difficult due to its enormous physical dimensions and the wide range of spatio-temporal scales involved in its dynamical behaviour. It has been recently proposed that the most efficient way to monitor the ocean is through networks of small, intelligent and cheap robotic platforms. Drifting profiling floats and gliders were developed in this context. Floats move with the currents meanwhile they periodically sample the water column through controlled immersions. Conversely, gliders are underwater autonomous vehicles with controllable motion at sea. Both platforms are extensively employed in oceanography due to their high autonomy. A network called Argo of around 3000 profiling floats spreads out around the world's ocean. Glider networks are starting to settle down at smaller scale in different places. The advent of these networks and the still scarce resources for ocean sampling, create a demand for quantitative tools for optimizing their use. In this work, the problem of optimally merging networks of profiling floats and gliders is considered. Specifically, a genetic algorithm is employed to find optimal glider trajectories to get together an unevenly distributed network of floats the best quality of the sampled field. A measure of the quality of the oceanographic field (objective function to minimize) is defined in terms of the mean formal error obtained from an optimum interpolation scheme. Results show that the quality of the sampled field can be greatly improved by merging both networks if the resolution of glider observations is adequately selected. The spatial lag between glider observations is related to the geometry of the network of profiling floats and must be order of the grid spacing obtained from the mean data spacing of the network of floats.

In the analysis of several scientific and engineering problems nonlinear dynamic systems are modelled as systems composed by the series connection of a linear dynamic subsystem with a nonlinear memoryless unit. If the nonlinearity follows the linear subsystem, the system is called of Wiener type, otherwise the system is called of Hammerstein type. In this work a nonparametric approach to the approximation of Wiener/Hammerstein models is proposed. The approach is based on the use of Laguerre filter banks to approximate the linear subsystem, and an artificial neural network to approximate the memoryless nonlinearity. Building on existing results of approximating properties of Laguerre filters and neural networks, theoretical convergence results of the approximating scheme to the underlining Hammerstein/Wiener model are reported. It is emphasized that the suggested approach requires much milder assumptions than those needed by other procedures previously proposed in the literature. In particular, no knowledge of the linear system order and time delay is needed, and the nonlinearity need not to be invertible and/or of polynomial type.

A two layered hierarchical architecture is considered as the fundamental scheme for the closed loop control of robots in operational space. By considering different kinds of information transfer from the outer to the inner controllers, it is shown that the architecture can take into account a wide class of control schemes, then the analysis of their stability and robustness properties can be performed in a unified framework. Then on this basis some general results concerning such properties are given.

In this work we propose a Lyapunov based design of velocity observers and the controller for stable trajectory tracking by a robotic manipulator. It is shown how the proposed design is exponentially stable over a finite domain, and, in the high gain approximation, exponentially stable in the large domain. Moreover, the design proposed leads naturally to a reduced order observer structure, with considerable implementation advantages.

This paper describes techniques and methodologies so far developed to investigate object fine-form discrimination by means of artificial tactile sensors. Sensor arrays, selectively sensitive to stress-tensor components and based on piezoelectric polymer technology, have been realized. Sensor output data are used to solve inverse elastic contact problems, by means of neural networks suitably trained to learn regularized inverse maps. Two possible neural network designs are considered: one is based on the multilayer perceptron trained with the standard backpropagation algorithm, and the other is based on the use of radial basis functions. In both cases, reconstruction of object shapes is demonstrated to be effective and robust with both simulated and real data.>

A class of coordinate transformations is introduced in order to derive the dynamic equation of a constrained manipulator on the constraining manifold. A theorem of existence is given, and some practical computational guidelines, based on the analytic knowledge of the equation of the constraining manifold, are outlined in order to compute the transformation off-line. If the analytic equations of the constraining manifold are inaccurate, the planned motion will not belong to the true manifold. It is then shown how the desired trajectories, both in the transformed domain and in the joint space domain, may be modified online to force a motion that belongs to the true manifold.>

The authors report the implementation and testing of a neural network algorithm to solve the fine-form discrimination problem. The skin-like sensor design is briefly reviewed, and the fine-form discrimination task is stated as an inverse problem of contact mechanics. The inverse problem is discussed using a singular value decomposition argument, to enlighten the basic regularity assumptions that are needed in the inversion, and the class of shapes that it is possible to recover from the data. The network design is given, along with the information about training and testing sets and learning rate. Simulated experimental results are reported.>

Learning control algorithms rely on offline estimates of generalized position, velocity, and acceleration errors. This problem is addressed in the framework of the spectral analysis of singular descriptor systems, and a class of robust estimates is introduced. The use of these estimators in the implementation of learning control algorithms has been tested on a prototype manipulator, and experimental results are reported. It is concluded that iterative learning control can be successfully implemented on real manipulators provided that robust estimators for velocity and acceleration are given.>

Techniques and methodologies being developed to investigate tactile fine-form discrimination through machine perception are described. A tactile sensor array capable of resolving the six independent components of the stress tensor field is briefly described. Robust numerical solutions of inverse elastic contact problems, through discrete inverse operators, are reported.>