This work proposes an approach for the fault detection (FD) of a single out-of-water ultrasound range meter for the measurement of sea height. Although they provide high-frequency and accurate measurements, such sensors are prone to faults, and the presence of foam on the water surface can hinder their measurement capabilities. In contrast, underwater pressure sensors provide more reliable but less accurate measurements of wave characteristics. Furthermore, wave attenuation below the sea surface affects the accuracy of pressure sensors-based measurements of wave motion. This work introduces a procedure for performing transfer learning, aimed at leveraging knowledge acquired from pressure sensor data to train a neural network that provides fault detection capabilities in ultrasonic measurement of sea level. The proposed transfer learning procedure takes advantage of linear wave theory, a description of the motion of small amplitude sea waves, to perform domain adaptation. The effectiveness of the proposed approach is discussed through simulations on a real-world dataset in which three common types of faults are injected.

This paper addresses the problem of feasible trajectory planning for an autonomous car-like vehicle. The proposed algorithm explicitly considers the closed-loop dynamics of the vehicle, implementation constraints and disturbance effects due to tyre-ground interaction. The trajectories are constructed incrementally as a sequence of connected segments. The feasibility of the trajectories is verified using a set-based approach. Finally, the effectiveness of the proposed approach is demonstrated by numerical simulations in two scenarios of interest for autonomous driving: self-parking and overtaking a slower vehicle, in which the algorithm proves to be able to find trajectories which can be tracked without violating constraints even in presence of different conditions of both the road and the vehicle.

This paper aims to propose a feasible solution to the trajectory planning problem for a constrained skid-steering mobile robot whose control system, sensors and actuators are connected through a communication network. Operating environment is firstly discretized by a finite dimensional grid. Then a weighted graph, whose nodes are the above mentioned points and whose arcs denote the trajectory segments connecting points is defined. Finally an algorithm to obtain the shortest feasible succession of segments connecting given starting and ending points is proposed. Trajectory feasibility is guaranteed in terms of sufficient conditions involving the solution of semi-definite programming (SDP) problems. In order to show the effectiveness of the proposed approach, some numerical simulations are proposed.