No abstract

No abstract

In the context of developing safe air transportation, our work is focused on understanding how Reinforcement Learning methods can improve the state of the art in traditional control, in nominal as well as non-nominal cases. The end goal is to train provably safe controllers, by improving both training and verification methods. In this paper, we explore this path for controlling the attitude of a quadcopter: we discuss theoretical as well as practical aspects of training neural nets for controlling a crazyflie 2.0 drone. In particular we describe thoroughly the choices in training algorithms, neural net architecture, hyperparameters, observation space etc. We also discuss the robustness of the obtained controllers, both to partial loss of power for one rotor and to wind gusts. Finally, we measure the performance of the approach by using a robust form of a signal temporal logic to quantitatively evaluate the vehicle's behavior. CCS CONCEPTS• Computer systems organization → Embedded and cyberphysical systems; Robotic control; • Software and its engineering → Formal methods; • Theory of computation → Modal and temporal logics; • Computing methodologies → Computational control theory; Reinforcement learning.

Abstract.To extend the functionalities of Advanced Driver Assistance Systems (ADAS) and have a more accurate control on the parameters of sensors mounted on an intelligent vehicle, a tool that can classify the scenarios which the vehicle moves in, is needed.This article presents a comparison of three classification techniques (PCA, ANN and SVM) to obtain a fast and robust scene classifier based only on images. The systems presented in this paper have been trained on three different categories of traffic scenarios: urban, highway, and rural, on a total of more than 23 hours of driving in different countries.

This paper presents the results obtained by the 2WIDE_SENSE Project, an EU funded project aimed at developing a low cost camera sensor able to acquire the full spectrum from the visible bandwidth to the Short Wave InfraRed one (from 400 to 1700 nm). Two specific applications have been evaluated, both related to the automotive field: one regarding the possibility of detecting icy and wet surfaces in front of the vehicle and the other regarding the pedestrian detection capability. The former application relies on the physical fact that water shows strong electromagnetic radiation absorption capabilities in the SWIR band around 1450 nm and thus an icy or wet pavement should be seen as dark; the latter is based on the observation that the amount of radiation in the SWIR band is quite high even at night and in case of poor weather conditions. Results show that even the use of SWIR and visible spectrum seems to be a promising approach; the use in outdoor environment is not always effective.