Resource-efficient real-time networking system for cyber-physical systems사이버-물리 시스템을 위한 자원 효율적인 실시간 네트워크 시스템 연구

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 353
  • Download : 0
Recent advances in embedded processors and software technologies have fostered the development of complex cyber-physical systems (CPS) such as autonomous driving vehicles. CPS typically rely on networks that interconnect sensors, controllers, and actuators to interact with physical worlds through real-time sensing and actuating. The last several years have seen a growing interest in developing networking technologies to provide deterministic and reliable communication, which is an essential requirement of safety-critical real-time applications in CPS. Beyond the essential requirement, another important requirement is efficient resource use. However, CPS networks often face inefficiency in resource use, because of the resource over-provisioning as follows: 1) pessimistic resource usage estimate for deterministic communication and 2) redundant transmission for reliable communication. To address such problems, many studies in the literature have proposed various resource-efficient real-time scheduling algorithms and fault management techniques, indicating that the adaptation is a key principle to deal with the problem. For each over-provisioning issue, some promising adaptation techniques have proposed including 1) mode-based scheduling and 2) on-demand fault recovery. However, applying those adaptation techniques into the CPS network yet raises a significant challenge. This is because it takes a long and unpredictable delay to dynamically update packet forwarding policy of the CPS network. Such a delay makes the networking system impossible to guarantee deterministic and reliable communication. This dissertation proposes a novel CPS networking system that effectively addresses the challenge to realize the adaptation approaches. To this end, we analyze delay factors of each adaptation step, then reduce and bound each delay factor by totally changing the way of adaptation. The proposed networking system consists of two parts: MC-SDN and RoD-SDN. MC-SDN realizes mode-based mixed-criticality scheduling which addresses the over-provisioning from the pessimistic resource usage estimate. It effectively supports mode-based dynamic packet scheduling with fast and predictable adaptation (i.e., mode change), and improves resource efficiency while preserving timing guarantee of real-time flows even in the worst case. RoD-SDN enables on-demand real-time fault recovery which addresses the over-provisioning due to the redundant transmission. With fast and predictable adaptation (i.e., path restoration), it can adaptively change the path of each flow to recover the communication from the link failure; it can improve resource utilization while providing reliability guarantee of real-time flows against network failure. We evaluated the proposed system with full implementation on top of a software switch (i.e., Open vSwitch) running on a network testbed composed of multiple single-board computers. In addition, we deployed the networking system into a 1/10 scale autonomous vehicle, and conducted case studies; the result shows that the proposed system improves resource efficiency while preserving system safety in a real-world application scenario.
Advisors
Shin, Insikresearcher신인식researcher
Description
한국과학기술원 :전산학부,
Publisher
한국과학기술원
Issue Date
2019
Identifier
325007
Language
eng
Description

학위논문(박사) - 한국과학기술원 : 전산학부, 2019.2,[vi, 73 p. :]

Keywords

Cyber-Physical Systems▼aReal-Time Network▼aMixed-Criticality Scheduling▼aFault Recovery System▼aSoftware-Defined Networking; 사이버-물리 시스템▼a실시간 네트워크▼a다중 중요도 스케줄링▼a고장 복구 시스템▼a소프트웨어-정의 네트워크

URI
http://hdl.handle.net/10203/265334
Link
http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=842407&flag=dissertation
Appears in Collection
CS-Theses_Ph.D.(박사논문)
Files in This Item
There are no files associated with this item.

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0