Rigorous design of autonomous systems for the Internet of Things (IoT)
- 講者Panagiotis Katsaros 教授 (School of Informatics, Aristotle University of Thessaloniki, Greece)
邀請人：陳郁方 - 時間2019-09-24 (Tue.) 10:00 ~ 12:00
- 地點資訊所新館101演講廳
摘要
AI-based autonomy is increasingly used for the automation of functions in smart systems (e.g. smart grids, smart transport systems), where various computational entities (edge/fog nodes, servers, clouds, applications and services) are tightly interacting with physical components through sensors and actuators. Such systems can accomplish tasks independently, or with minimal supervision from humans and can optimise the management of resources, in complex and unpredictable environments. Let us consider the case of sensor devices situated in heterogeneous environments that change over time e.g. due to sensors mobility or changes in the weather conditions. Instead of looking at these devices as energy-constrained sources of data, to avoid both over-dimensioning them with larger than necessary energy sources and risking with less frequent than necessary sensing (in order to save energy), we see them as autonomous agents with adaptive sensing capabilities.In this seminar, I will first present my current research work on the rigorous model-based design of IoT systems. This is only the first step towards our vision for the rigorous design of smart autonomous systems that will be trustworthy enough for their intended use. Serious concerns remain about the trustworthiness and safety of current machine learning technologies, whereas formal model-based design approaches are generally applicable only to non-learning systems that operate in well-characterised environments. We aim to
• combine the best of model-based design with formal guarantees
• leverage the advances of machine learning and
• introduce novel machine learning and synthesis techniques, developed specifically for system design.