Solutions for machine learning and decision on autonomous platform in a real-time context
Offer N°: 14160
The human brain can realize high-level tasks, such as pattern recognition, in real-time, with a low energy consumption and a non-extensible volume. Nevertheless, even supposedly powerful hardware architectures cannot fully address these problems that remain true technological challenges, despite an extensive literature.
Study and Design of a data memory scheme for embedded vision systems
Offer N°: 14158
The embedded vision systems are growing rapidly due to evolution of microelectronics technologies and are used in many fields (robotics, automation, personal assistance, industrial control, driving assistance, surveillance). These evolutions allow the miniaturization of vision systems and provide the opportunity of major changes in their architecture.
Hybrid memory hierarchy and dynamic data handling for embedded multicore architectures
Offer N°: 14156
Due to technological limitations on CMOS operating frequency, the race for ever increasing processing performance in embedded systems today translates to a rapid increase of the number of cores per device. But the over-proportional raise of power consumption is increasingly hindering this performance race.
Multi strained channels integration for 10nm FDSOI CMOS
Offer N°: 14150
This PhD thesis is focused on the evaluation of different strain techniques for boosting the electrical performance of FDSOI CMOS for sub 10nm generation nodes.