The VI-DAS proposal has been granted by the EU Commission. It focuses on highly automated driving, specifically on switching between different automation levels and the effect this has on drivers. In this project, MPS will focus on the environment perception technologies that are required for safe and comfortable automated driving.
Our research cluster has a new member: Anweshan Das. He started his PhD research on the topic of accurate localization for highly automated vehicles. This research is performed in a joint program together with TomTom and other EU partners.
Our paper "FCNs for Free-Space Detection with Self-Supervised Online Training" of Willem Sanberg and Gijs Dubbelman, has been accepted for publication at the international workshop on Deep-driving: learning representations for intelligent vehicles workshop at the IEEE Intelligent Vehicles Symposium. It will be presented on the 19th of June 2016 in Gothenburgh Sweden.
The mobile perception research cluster focuses on technologies that allow moving sensor platforms to perceive their environment. MPS is part of the Signal Processing Systems group of TUE's electrical engineering department and specifically of its VCA research group. We also work closely with the mechanical engineering department's Dynamics and Control and Control Systems Technology groups. The application domains that we currently focus on are automotive, transportation, and defence. Some of our key industrial partners include: TomTom, Mapscape, NXP, Ford, TASS, and NVidia.
Key research domains of MPS are:
Machine learning and pattern recognition:
In short, machine learning and pattern recognition deal with extracting meaningful information from sensor data. We specifically focus on methods that can be used in real-time in the near future (5 to 10 years). Currently, a very promising group of techniques are that of deep learning using artificial neural networks.
3D computer vision:
Computer vision researches pattern recognition and machine learning techniques specifically for visual data. We also focus on 3D computer vision techniques that allow estimating 3D geometric information from visual data. This is important for mapping and localization applications.
How to combine and fuse data from different sources in order to improve accuracy and reliability of environment perception. For this, we work work with typical and next-generation automotive sensors like RADAR, LIDAR, LEDDAR bit also data derived from I2V communication.