期刊
JOURNAL OF SYSTEMS ARCHITECTURE
卷 107, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.sysarc.2020.101736
关键词
Artificial DNA; Artificial hormone system; Self-organization; Dependability; Dynamic merger and separation
In the last decade, bio-inspired techniques like self-organization and emergence have been in the focus of several research projects to deal with the challenge to develop, to configure and to maintain highly distributed and embedded systems. They are promising approaches to handle real-world applications such as control tasks in cars which may suffer from processor or task failures and must adapt to a changing environment. In biology the structure and organization of a system is coded in its DNA, while dynamic control flows are regulated by the hormone system. We adapted these concepts to embedded systems using an artificial DNA (ADNA) and an artificial hormone system (AHS) and both were implemented in a middleware. Based on these concepts, highly reliable, robust and flexible systems can be created. These properties predestine the ADNA and AHS for the use in future automotive applications. We showed in recent publications several examples for the use of the ADNA/AHS approach dealing with automotive applications running on the processors of a distributed system. In this contribution, we compute the failure probabilities of a p: 1 redundant system where a group of p processors shares a redundant processor and compare it with the failure probability of the ADANA system. The results motivate us to go one step further and to extend the ADNA/AHS approach for applications running on the processors of different distributed systems which normally operate as isolated systems, but may temporarily merge if they get in communication range. We show this in an automotive example where a driving car uses the processors of others cars to run its tasks. We also evaluate performance measures like down-times, reaction times and so on. The evaluations show that this approach is very promising for automotive applications.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据