26th International Conference on Optical Fiber Sensors, 2018
We present, for the first time, a colour coded sequence running in an endless loop and dedicated ... more We present, for the first time, a colour coded sequence running in an endless loop and dedicated to ultra-fast Brillouin distributed sensors. Measurements within the whole elastic zone of single-mode optical fibers are presented.
Autonomous navigation vehicles have rapidly improved thanks to the breakthroughs of Deep Learning... more Autonomous navigation vehicles have rapidly improved thanks to the breakthroughs of Deep Learning. However, scaling autonomous driving to low-power and real-time systems deployed on dynamic environments poses several challenges that prevent their adoption. In this work, we show an end-to-end integration of data, algorithms, and deployment tools that enables the deployment of a family of tiny-CNNs on extra-low-power MCUs for autonomous driving mini-vehicles (image classification task). Our end-to-end environment enables a closed-loop learning system that allows the CNNs (learners) to learn through demonstration by imitating the original computer-vision algorithm (teacher) while doubling the throughput. Thereby, our CNNs gain robustness to lighting conditions and increase their accuracy up to 20% when deployed in the most challenging setup with a very fast-rate camera. Further, we leverage GAP8, a parallel ultra-low-power RISC-V SoC, to meet the real-time requirements. When running a ...
Interferometric measurements beyond the coherence length of the laser are investigated theoretica... more Interferometric measurements beyond the coherence length of the laser are investigated theoretically and experimentally in this paper. Thanks to a high-bandwidth detection, high-speed digitizers and a fast digital signal processing, we have demonstrated that the limit of the coherence length can be overcome. Theoretically, the maximal measurable displacement is infinite provided that the sampling rate is sufficiently short to prevent any phase unwrapping error. We could verify experimentally this concept using a miniature interferometer prototype, based on a frequency stabilized vertical cavity surface emitting laser. Displacement measurements at optical path differences up to 36 m could be realized with a relative stability better than 0.1 ppm, although the coherence length estimated from the linewidth and frequency noise measurements do not exceed 6.6 m.
ABSTRACT Recently advances in the design of general purpose SoC have revolutionized the conceptio... more ABSTRACT Recently advances in the design of general purpose SoC have revolutionized the conception of embedded systems, thanks to the availability of several, most of the time asymmetric, units of calculation. Nevertheless a deep knowledge of the internal hardware architecture of these multi-core SoC is required to efficiently benefit from its large computing performance. The current work, the SOSoC project, provides a Linux framework which offers the opportunity to application developers to optimize in a transparent way the execution of algorithms by dispatching tasks, dynamically or statically, among the various available cores to benefit a maximum from the available hardware.
26th International Conference on Optical Fiber Sensors, 2018
We present, for the first time, a colour coded sequence running in an endless loop and dedicated ... more We present, for the first time, a colour coded sequence running in an endless loop and dedicated to ultra-fast Brillouin distributed sensors. Measurements within the whole elastic zone of single-mode optical fibers are presented.
Autonomous navigation vehicles have rapidly improved thanks to the breakthroughs of Deep Learning... more Autonomous navigation vehicles have rapidly improved thanks to the breakthroughs of Deep Learning. However, scaling autonomous driving to low-power and real-time systems deployed on dynamic environments poses several challenges that prevent their adoption. In this work, we show an end-to-end integration of data, algorithms, and deployment tools that enables the deployment of a family of tiny-CNNs on extra-low-power MCUs for autonomous driving mini-vehicles (image classification task). Our end-to-end environment enables a closed-loop learning system that allows the CNNs (learners) to learn through demonstration by imitating the original computer-vision algorithm (teacher) while doubling the throughput. Thereby, our CNNs gain robustness to lighting conditions and increase their accuracy up to 20% when deployed in the most challenging setup with a very fast-rate camera. Further, we leverage GAP8, a parallel ultra-low-power RISC-V SoC, to meet the real-time requirements. When running a ...
Interferometric measurements beyond the coherence length of the laser are investigated theoretica... more Interferometric measurements beyond the coherence length of the laser are investigated theoretically and experimentally in this paper. Thanks to a high-bandwidth detection, high-speed digitizers and a fast digital signal processing, we have demonstrated that the limit of the coherence length can be overcome. Theoretically, the maximal measurable displacement is infinite provided that the sampling rate is sufficiently short to prevent any phase unwrapping error. We could verify experimentally this concept using a miniature interferometer prototype, based on a frequency stabilized vertical cavity surface emitting laser. Displacement measurements at optical path differences up to 36 m could be realized with a relative stability better than 0.1 ppm, although the coherence length estimated from the linewidth and frequency noise measurements do not exceed 6.6 m.
ABSTRACT Recently advances in the design of general purpose SoC have revolutionized the conceptio... more ABSTRACT Recently advances in the design of general purpose SoC have revolutionized the conception of embedded systems, thanks to the availability of several, most of the time asymmetric, units of calculation. Nevertheless a deep knowledge of the internal hardware architecture of these multi-core SoC is required to efficiently benefit from its large computing performance. The current work, the SOSoC project, provides a Linux framework which offers the opportunity to application developers to optimize in a transparent way the execution of algorithms by dispatching tasks, dynamically or statically, among the various available cores to benefit a maximum from the available hardware.
Uploads