Post-Position: Neural Networks and Heterogeneous Architectures for Multi-Sensor Systems in Autonomous Cars
Neural Networks and Heterogeneous Architectures for Multi-Sensor Systems in Autonomous Cars
LAMIH/CNRS, University of Valenciennes, France
Progresses in the design of CMOS circuits have made the possibility to support very complex Machine Learning (ML) algorithms using large data sets. For this reason, AI techniques such as Convolution Neural Network (CNN) and Deep Neural Network (DNN) have received recently interests both in industry and academy to implement complex applications.
In the domain of embedded systems for automotive applications, CNN and DNN have many potential applications, especially for Advanced Driving Assistance Systems (ADAS) and autonomous driving. These algorithms have shown high performances in scene understanding and object/obstacle classification.
This post-doc aims to contribute in the domain of embedded system design for automotive applications, especially in autonomous driving. The objective here is to develop new heterogeneous Multicore/FPGA/GPU-based architectures to support complex ML algorithms. The target architectures must in one hand adapt the ML algorithm and the supporting architecture to the different scenarios and on the other hand must use different CNN and DNN configurations taking into account the different characteristics of embedded sensors (Cameras, Lidars, Radars).
The duties also include collaboration with PhD students working on these topics and helping to write high-impact papers and funding applications.
Bibliography:
- Design of Multiple-Target Tracking System on Heterogeneous System-on-Chip Devices, G. Zhong, S.Niar, A.Prakash, T.Mitra, IEEE Trans. Vehicular Technology 65(6), 2016.
- An Accelerator for High Efficient Vision Processing, Z. Du, S.Liu, R.Fasthuber, T. Chen, P. Ienne, L. Li, T. Luo, Q. Guo, X. Feng, Y. Chen, and O. Temam, IEEE Transactions on CAD of Integrated Circuits and Systems, 02/2017
- Radar signature in multiple target tracking system for driver assistant application, H. Liu, S. Niar, IEEE/ACM DATE 2013.Computer Vision for Autonomous Vehicles, J.Janai, F. Guney, A.Behl, A. Geiger, Datasets and State-of-the-Art. CoRR, 2017.
Required degree and skills:
- Ph.D in computer engineering/electrical engineering/automation.
- Experience in scientific journals / conference publication with good English (writing and speaking).
- Knowledge/experience in one of the following matters would be an advantage: 1 Machine learning and AI techniques, 2 Signal and/or image processing, 3 Embedded FPGA-GPU systems, hardware architectures.
An application prepared in English or French should contain:
- CV with the list of publications.
- Contact information for 2 reference persons.
Salary:
2500 euros/month; Deadline: 30/10/2017; Duration: 12 months + 6 months; Preferred starting date: 01/10/2017 but not later than 01/12/2017
Contact:
Professor Smail NIAR Smail.niar@univ-valenciennes.fr LAMIH/CNRS - University of Valenciennes, France. www.univ-valenciennes.fr/LAMIH/membres/niar_smail