SAGE-CPSoC · Self-Aware CPSoCs with Hierarchical Goal Management

Self-Aware CPSoCs with Hierarchical Goal Management

Advanced computing and automated devices have enhanced human life style by a great deal, providing a wide range of value and services in every facet of the society. A wide spectrum of computing paradigms exist covering massively parallel data centres, complex Cyber-Physical Systems, handheld mobile embedded systems till ultra-low power embedded devices, also termed Internet-of-things. Compute platforms are responsible for real world data acquisition through sensors, efficient processing of this data, intelligent decision making or actuation based on processing and finally for establishing robust communication and networking among several modules. Design of efficient computing platforms is essential for realizing novel future applications and workloads, ensuring high performance and quality of service at lower power consumption and cost. However, due to diverse workload characteristics, designers are forced to develop highly customized processing units at the loss of generality or highly generic systems with conflicting and orthogonal constraints. With this in view, designing smarter systems that can adapt to varying application requirements through learning strategies and high autonomy becomes necessary. This project aims at empowering autonomy through Self-awareness in computing systems. Self-awareness helps systems to understand, manage, and report on their own system behavior. In addition to its roots in psychology, the notion of self-awareness has been used in computing in a variety of different domains such as autonomic computing, organic computing, adaptive systems, and self-organizing systems. Self-awareness allows a system to deal better with complexity. The complexity comes from the system itself, from the environment, and from the exceedingly diverse goals and objectives it has to meet. As a key component of self-awareness, this project leverages goal-driven autonomy (GDA) for goal management. Cognitively, humans pursue number of concurrent goals - spawning, terminating, (re-)prioritizing them dynamically by arranging them in hierarchical structures. GDA is a goal reasoning model which allows agents to dynamically generate their goals in response to environmental changes. For this reason, this project proposes for resource allocation and management in computing systems. The overall objectives in this project are: • To identify the key properties and functions for the hierarchical goal management (HGM) capability • To adapt and integrate design methods, algorithms on self-awareness and online learning to the system • To adapt and integrate the state-of-the-art management policies to the CPSoC platform to be orchestrated and managed hierarchically With 20-40 Billion on-line devices to be expected within the next decade, they have to be robust, dependable, and adaptive, with high utility and efficiency in changing environmental conditions. Self-aware, adaptive artificial subjects, as envisioned in this project, will continuously find new, useful applications in domains like medicine, transport, food security, environmental monitoring, surveillance, etc.

Текст от CORDIS, на английски · Данни: CORDIS, © Европейски съюз