Matteo Cavalletti

Lyapunov-based switching control using neural networks for a remotely operated vehicle

This paper considers the tracking control problem of an underwater vehicle subjected to different load configurations, which from time to time introduce considerable variations of its mass and inertial parameters. The control of this kind of mode-switch process cannot be adequately faced with traditional adaptive control techniques because of the too long time needed for adaptation. To cope with this problem, a switching control scheme is proposed and the stability of this multi-controller system is analysed using the Lyapunov theory. The performance of the switched controller is evaluated by numerical simulations.

Online tuned neural networks for fuzzy supervisory control of pv-battery systems

The paper deals with a neural network based fuzzy supervisor control to manage power flows in a Photo-Voltaic (PV) - Battery system. An on-line self-learning prediction algorithm is used to forecast, over a determined time horizon, the power mismatch between PV production and electrical consumptions. The learning algorithm is based on a Radial Basis Function (RBF) network and combines the growing criterion and the pruning strategy of the minimal resource allocating network technique. The power flows are scheduled by a Fuzzy Logic Supervisor (FLS) which controls the charge and discharge of a battery used as an energy buffer. The proposed solution has been experimentally tested on a 14 KWp PV plant and a lithium battery pack.

Intelligent control for a remotely operated vehicle

This article considers the tracking control problem of an underwater vehicle used in the exploitation of combustible gas deposits at great sea depths. The vehicle is subjected to different load configurations that introduce considerable variations of its mass and inertial parameters. In this work it is assumed that the possible vehicle configurations are known, but the time instants when the changes occur and the new vehicle configuration following the change are unknown. A neural network-based switching control is proposed for the considered mode-switch process. This solution simplifies the control scheme implementation and reduces the control signal chattering.

Solar irradiation forecasting for PV systems by fully tuned minimal RBF neural networks

An on-line prediction algorithm able to estimate, over a determined time horizon, the solar irradiation of a specific site is considered. The learning algorithm is based on Radial Basis Function (RBF) networks and combines the growing criterion and the pruning strategy of the minimal resource allocating network technique. An adaptive extended Kalman filter is used to update all the parameters of the Neural Network (NN). The on-line learning mechanism avoids the initial training of the NN with a large data set. The proposed solution has been experimentally tested on a 14 kWp PhotoVoltaic (PV) plant and results are compared to a classical RBF neural network.

Switching Adaptive Control for a Remotely Operated Vehicle

This paper considers the tracking control problem for an underwater vehicle subjected to different load configurations, that result in considerable variations of its mass and inertial parameters. Depending on the operative situation, the different possible vehicle configurations could not be known in advance. In general, it is not a priori known neither when the operating conditions are changed nor which is the new vehicle configuration after the change. The control of this class of mode-switch processes can not be adequately faced with traditional adaptive control techniques because of the too long time needed for adaptation when a change of the configuration occurs. To cope with this problem, an adaptive control strategy endowed with a supervised switching logic is proposed. This strategy is particularly suitable when the different possible vehicle configurations are not a priori known. The stability of this switched control system is analyzed using the Lyapunov theory and its performance is evaluated by numerical simulations.

ADAPTIVE NONLINEAR CONTROL FOR VESSEL ROLL STABILIZATION

Abstract The paper describes an adaptive neural network based nonlinear control for roll stabilization of ocean motoryachts. The advantage of the adaptive controller is to improve vessel roll stabilization performance under a variety of operational and environment conditions. The application of nonlinear control theory allows to compensate for nonlinearities in the control design. The analysis of the control stability is based on the Lyapunov theory. The proposed adaptive controller is tested by a simulation study based on a nonlinear model, describing the dynamics of a vessel in four degrees of freedom. The wave excitation forces are simulated as multisine time series.

Lyapunov-Based Switching Control for a Remotely Operated Vehicle

This paper considers the tracking control problem of an underwater vehicle subjected to different load configurations, which from time to time introduce considerable variations of its mass and inertial parameters. The control of this kind of mode-switch process cannot be adequately faced with traditional adaptive control techniques because of the too long time needed for adaptation. To cope with this problem, a switching control scheme is proposed and the stability of this multi-controller system is analyzed using the Lyapunov theory. The performance of the switched controller is evaluated by numerical simulations

Neural network-based switching adaptive control for a remotely operated vehicle

Underwater vehicles operate in dynamic environments where sudden changes of the working conditions occur from time to time. The need of an effective control action calls for refined techniques with a high degree of robustness with respect to large parametric variations and/or uncertainties. Supervised switching control seems to be theoretical frameworks where appropriate control strategies can be developed. In this chapter, a switching adaptive tracking control based on NNs is proposed and compared with an NN-based switching controller. The form of used nets is the RBFN, which has been used successfully in other control system applications (Antonini et al., 2006) and has favourable characteristics in terms of the best approximation property (Poggio & Girosi, 1990). On the basis of numerical tests, the NNSAC is able to cope with the large transient errors related to the considered mode-switch processes, when knowledge of the different possible vehicle configurations is poor. In fact, if the operative conditions are unknown, an NNSC cannot guarantee good control performance; the pre-computed controllers cannot cope with all environment and load conditions. Therefore, the integration of a switching control strategy with adaptive controllers is particularly well suited to cope with these unknown operative conditions and improve the performance of the overall control system when the different environments where the vehicle operates are not well known.

Application of a Supervised Improved PID for the scheduling of energy feeding in a PV - battery system

Abstract This paper deals with the development of a PID control architecture for better utilization of the storage battery connected to a PhotoVoltaic (PV) Plant. The problem of the stochastic nature of the PV plant is overcomed scheduling the power feeding of the electric line. A neural network is used to derive the one-day-forecast of the PV production and a supervised PID controller is proposed to control the charge and discharge current reference to the battery, that is used as an energy buffer. The communication between all the parts of the system and the supervisor controller is made via TCP/IP protocol. The Energy Resources company has supported the experimentally tests of the proposed solution on a 14 KWp PV plant and a lithium battery pack.

Statistical learning controller for the energy management in a fuel cell electric vehicle

This paper considers a high efficiency energy management control strategy for a hybrid fuel cell vehicle. The proposed switching architecture consists of a bank of neural network based controllers designed using statistical learning theory. The use of different power sources and the presence of different constraints make the power management problem highly nonlinear. Probabilistic and statistical learning methods are used to design the weights of a neural network and the switching strategy is used to implement different controllers designed on the considered operative conditions. The proposed controller increases the efficiency of the whole system and reduces the fuel consumption during a given path. Numerical results are obtained using the model of a real hybrid car, ¿smile¿ developed by FAAM, using a stack of fuel cells as the primary power source in addition to ultracapacitors and a lithium battery pack. The results are compared with those of a single neural network based controller and the performance is shown to be satisfactory in terms of fuel consumption and the efficiency of the whole system.