Phd Position at Ietr - Rennes, France - INSTITUT NATIONAL DES SCIENCES APPLIQUEES

Description

PhD Position at IETR

• Ecodesign of Embedded Systems applied to Smart Vision

:

• Réf
  ABG-123956
• Sujet de Thèse 17/05/2024
• Contrat doctoral
• INSTITUT NATIONAL DES SCIENCES APPLIQUEES
• Lieu de travail
• Rennes
• Bretagne
• France
• Intitulé du sujet
• PhD Position at IETR
• Ecodesign of Embedded Systems applied to Smart Vision
• Champs scientifiques
• Electronique
• Energie

Description du sujet:


Context and Challenge

Like all industrial sectors, the electronics sector must, to be sustainable, drastically reduce its CO2 emissions by 2050, and more globally commit to a logic of circular economy.

While environmental impacts of electronics are increasingly understood, the evaluation of impacts in Life Cycle Assessments (LCA) is still an emerging practice that requires complex models, data and insights.

Once the impacts are known, ecodesign strategies must be developed and assessed.

The impacts of an electronic system arise from its manufacturing, use and end-of-life.

The system itself is composed of printed circuit boards (PCB), semiconductor integrated circuits (ICs), passive components, connectors, sensors, actuators, batteries and displays, each having environmental impacts.

In particular, the greenhouse gas (GHG) emissions of semiconductor manufacturing represent an important share of the embodied carbon footprint of electronics systems and must be considered alongside energy consumption and end-of-life [M21][P22][U22].

A smart vision system is a combination of sensors and digital processing that captures, preprocesses and distributes a video or a semantic description of a visual scene.

It can take the form of a smart camera, mono-sensor [B14] or multi-sensors [K07], or of a camera module in an embedded system comprising a sophisticated image processing pipeline [M20].

It can also comprise a set of distributed sensors [R10].

Current smart vision systems embed optics, Complementary Metal-Oxide-Semiconductor (CMOS) sensors, color processing, image enhancement, video compression and artificial intelligence for image and video analysis.

The sophistication of these systems is increasing, as examplified in smartphones that combine several sensors, dedicated Image Signal Processors (ISPs) and software post-processing to implement computational photography [D21].

This thesis will explore ecodesign of smart vision systems as representative examples of high performance embedded systems. The carbon impact of such systems is currently dominated by the embodied emissions, i.e.

the greenhouse gas emissions of system fabrication [G21], but results from a complex and difficult to evaluate combination of fabrication, use and end-of-life impacts of its components and software layers [G21] as illustrated in Figure 1.

The thesis will aim at understanding how to model linear and non-linear factors influencing carbon emissions of such a system and developing ecodesign methods to minimize environmental impacts by leveraging on circular economy.

Objectives

• Novel methods for life cycle analysis of embedded systems. System ecodesign requires precise information of the impacts of system design decisions (types of CPU/GPU, FPGA, memory, busses, connectors, PCBs, etc.) and on the use phase of the system (idle time, workload, etc.). Thus, the first objective of this thesis will be to improve Life Cycle Analyses (LCAs) of embedded systems, especially applied to carbon emissions evaluation of smart vision systems. Different forms of LCAs exist that attribute costs to subsystems and that evaluate costs of modifications. However, environmental impacts of embedded systems are currently very partially known, notably due to lack of manufacturer data and to excessive approximations. In this thesis, new forms of LCA will be designed, adapted to the specific context of smart vision systems, and exploiting the research teams expertise in system design. In particular, real-life smart camera cases will be analyzed, prototyped, and measured for energy consumption, based on the experience of Institut Pascal in smart camera prototype design.
• Ecodesign for reducing carbon impacts. A second objective will be to analyse the options for modifying such systems towards a more sustainable form while preserving performances. Ecodesign means designing the embedded system with the goal to be part of circular economy, i.e. decreasing the need of new materials, increasing reuse, and avoiding waste. Novel scene observation strategies will be analyzed as well as system downsizing, remanufacture [K06], reuse of recovered or low impact sensors or processing devices, improve repairability, and trade-offs between processing and sensing costs. This study will aim at proposing novel hardware design methods that, for a given level of system service, provably reduce environmental impacts.

Environment and Expected Impact
The thesis will be co-directed by Dr. Maxime Pelcat and Prof. François Berry, and advised by Dr. Thibaut Marty.