The deployment of Artificial Intelligence (AI) based solutions to deliver advanced software functionalities is consolidating as a key competitive factor in several industrial domains. In the automotive industry, for instance, autonomous driving (AD) software is meant to support autonomous operation and decision-making for all aspects in a vehicle, by processing a massive amount of data coming from multiple sensors like cameras and LiDARs. The entailed computational requirements can only be matched by complex MPSoCs (Multi-Processor System on Chip) with generic and ad-hoc hardware accelerators. Moreover, the increasing complexity of AI-based software functionalities encourages the use of highly modular middleware frameworks such as ROS2, CyberRT, or Autoware, running on top of general-purpose and/or real-time operating systems. Performance, resilience to hardware faults, and (timing) analyzability are fundamental (and sometimes conflicting) requirements for this type of system, where extensive guarantees must be provided on the capability to deliver correct results in a timely manner, as dictated by domain-specific Functional Safety (FuSa) standards.
We are seeking for highly-motivated, brilliant candidates to enroll in a PhD position in the area of performance optimization, fault-tolerance, and analysability of system software for AI-driven critical systems. System software, including (Real-time)OS and hypervisors, low-level run-time and AI libraries, play a key role in ensuring that those hardware and software functionalities that can concur in achieving those (possibly conflicting) requirements are properly and consistently configured, and not jeopardized by the system software at operation. The candidate is expected to master at the end of the PhD existing approaches and elaborate novel solutions to extend and improve software level support to achieve high performance, analysability, and FuSa compliance in general. The PhD will particularly focus on the requirements stemming from complex, functionally rich AI-based applications, characterized by tight interaction and strong dependences between hardware layer, system software, AI libraries, and (user-level) functions. The candiate at the end of the PhD is meant to understand and model those interactions aiming to maximize performance, timing analyzability, and FuSA aspects, including resilience to hardware faults. Hence, the candidate is expected to combine software analysis, applied investigation, and hands-on implementation of system software to analyze, model, and optimize the complex interaction between hardware, low-level software, AI modules, and other elements in the software stack.
Key Duties
- Familiarization with existing system software solutions and research proposals in critical embedded systems to improve time analyzability, resilience to hardware faults, and performance
- Design or tailor RTOS and hypervisor design to support AI-based autonomous driving functionalities achieving the required levels of performance, timing analyzability, and resilience to hardware faults
- Familiarization with representative AI-based autonomous driving setup and simulation
- Identify and explore opportunities for optimization of RTOS/hypervisor configuration, platform configurations and AI-models to meet performance requirements
Data de tancament: Divendres, 15 Novembre, 2024
Més posts de Recerca
Cap comentari:
Publica un comentari a l'entrada