Please use this identifier to cite or link to this item: https://www.um.edu.mt/library/oar/handle/123456789/28186
Title: Run-time monitoring for the diagnosis and recovery of complex physical systems
Authors: Cataldo, Giacinto De
Colombo, Christian
Franco, Antonio
Pace, Gordon J.
Valentino, Gianluca
Vella, Kevin
Keywords: Computer science
Computer software -- Verification
Intelligent control systems
Issue Date: 2016-06
Publisher: ATTRACT
Citation: De Cataldo, G., Colombo, C., Franco, A., Pace, G. J., Valentino, G., & Vella, K. (2016). Run-time monitoring for the diagnosis and recovery of complex physical systems. Trends, Wishes and Dreams Symposium on Detection and Imaging Technologies, Barcelona.
Abstract: Runtime verification focuses on techniques to check the dynamic (runtime) behaviour of a system typically with the aim of ensuring that the system is working correctly. In lining property checks via assertions or similar techniques has been standard practice since the dawn of programmable machines. Such in lined approaches result in the interweaving of the executable system specification (the program) with the specification of properties which it should satisfy. Separation of concerns has long been identified as an important principle, and the possibility of separating these two aspects of a system is one of the objectives of most modern runtime verification approaches — allowing for (i) having different teams working on the different aspects i.e. development and quality assurance; (ii) the use of the specification across different versions, instances or even systems. Another issue with in lined assertion checking arises as the complexity of the properties increases. Although in lining a property such as ‘the gas leak variable should be low when the induce Spark method is called’ is straightforward, a property such as ‘the open Valve method should have been called before induce Spark is called’ results with the developer having to introduce additional state to remember whether open Valve was called in the past. More complex properties, such as ‘The gas leak variable may not have been true for more than 1 minute in the last 30 minutes just before induce Spark is called’, results in more complex additional state and logic to handle it, which may, in turn result in more new errors being introduced into the system. The role of the runtime verification tool is twofold: (i) it modifies the system instrumenting code to be able to capture points of interest during its execution which are of interest with respect to the specification; and (ii) it converts the specification into a monitor, which reacts whenever a point of interest is reached, checking that the behaviour of the system does not violate the specification. Using such an approach, a specialised language can be used to write the specifications, which allows the adoption of domain specific languages which can be used to describe the behaviour more succinctly and precisely. Runtime verification would lend itself well to complex and critical environments such as detector and accelerator control systems, where loss of detector sub-systems could hinder it from data taking, or errors in control systems of high-energy particle accelerators with highly-destructive beams such as LHC could damage the machine, leading to months of costly downtime.
URI: https://www.um.edu.mt/library/oar//handle/123456789/28186
Appears in Collections:Scholarly Works - FacICTCS

Files in This Item:
File Description SizeFormat 
2016 - TWD - RuntimeMonitoring.pdf
  Restricted Access
570.13 kBAdobe PDFView/Open Request a copy


Items in OAR@UM are protected by copyright, with all rights reserved, unless otherwise indicated.