Reinforcement learning is considered to be a strong AI paradigm which can be used to teach machines through interaction with the environment and learning from their mistakes. Despite its perceived utility, it has not yet been successfully applied in automotive applications. Motivated by the successful demonstrations of learning of Atari games and Go by Google DeepMind, we propose a framework for autonomous driving using deep reinforcement learning. This is of particular relevance as it is difficult to pose autonomous driving as a supervised learning problem due to strong interactions with the environment including other vehicles, pedestrians and roadworks. As it is a relatively new area of research for autonomous driving, we provide a short overview of deep reinforcement learning and then describe our proposed framework. It incorporates Recurrent Neural Networks for information integration, enabling the car to handle partially observable scenarios. It also integrates the recent work on attention models to focus on relevant information, thereby reducing the computational complexity for deployment on embedded hardware. The framework was tested in an open source 3D car racing simulator called TORCS. Our simulation results demonstrate learning of autonomous maneuvering in a scenario of complex road curvatures and simple interaction of other vehicles.

We present research using the latest reinforcement learning algorithm for end-to-end driving without any mediated perception (object recognition, scene understanding). The newly proposed reward and learning strategies lead together to faster convergence and more robust driving using only RGB image from a forward facing camera. An Asynchronous Actor Critic (A3C) framework is used to learn the car control in a physically and graphically realistic rally game, with the agents evolving simultaneously on tracks with a variety of road structures (turns, hills), graphics (seasons, location) and physics (road adherence). A thorough evaluation is conducted and generalization is proven on unseen tracks and using legal speed limits. Open loop tests on real sequences of images show some domain adaption capability of our method.

We present the first far infrared (FIR) dust emission polarization map covering the full extent Milky Way's Central molecular zone (CMZ). The data, obtained with the PILOT balloon-borne experiment, covers the Galactic Center region −2 • < < 2 • , −4 • < b < 3 • at a wavelength of 240 µm and an angular resolution 2.2 . From our measured dust polarization angles, we infer a magnetic field orientation projected onto the plane of the sky that is remarkably ordered over the full extent of the CMZ, with an average tilt angle of 22 • clockwise with respect to the Galactic plane. Our results confirm previous claims that the field traced by dust polarized emission is oriented nearly orthogonal to the field traced by GHz radio synchrotron emission in the Galactic Center region. The observed field structure is globally compatible with the latest Planck polarization data at 353 GHz and 217 GHz. Upon subtraction of the extended emission in our data, the mean field orientation that we obtain shows good agreement with the mean field orientation measured at higher angular resolution by the JCMT within the 20 km/s and 50 km/s molecular clouds. We find no evidence that the magnetic field orientation is related to the 100 pc twisted ring structure within the CMZ. We propose that the low polarization fraction in the Galactic Center region and the highly ordered projected field orientation can be reconciled if the field is strong, with a 3D geometry that is is mostly oriented 15 • with respect to the line-of-sight towards the Galactic center. Assuming equipartition between the magnetic pressure and ram pressure, we obtain magnetic field strengths estimates as high as a few mG for several CMZ molecular clouds.

No abstract

Future cosmology space missions will concentrate on measuring the polarization of the Cosmic Microwave Background , which potentially carries invaluable information about the earliest phases of the evolution of our universe. Such ambitious projects will ultimately be limited by the sensitivity of the instrument and by the accuracy at which polarized foreground emission from our own Galaxy can be subtracted out. We present the PILOT balloon project which will aim at characterizing one of these foreground sources, the polarization of the dust continuum emission in the diffuse interstellar medium. The PILOT experiment will also constitute a test-bed for using multiplexed bolometer arrays for polarization measurements. We present the results of ground tests obtained just before the first flight of the instrument.