School of EEITM
University of Surrey
Guildford, Surrey
GU2 5XH, UK
Tel: +44 (0)1483 259823
Fax: +44 (0)1483 876051
|
|
School of EEITM University of Surrey Guildford, Surrey GU2 5XH, UK |
Tel: +44 (0)1483 259823 Fax: +44 (0)1483 876051 |
Since the production of the last newsletter there have been two meetings of the Safety-Critical Systems Round Table, both held at the University of Surrey. The meeting in December served to delimit particular aspects of modelling and design which must be attended to in ensuring safe practice when using simulation software. The members of the Round Table engaged in a brainstorming session followed by a discussion and review of the emerging structure describing the modelling process. This structure, in terms of stages and associated hazards was elaborated and refined by means of a postal questionnaire after the meeting.
The structure thus formed became the basis of the meeting in March where the Round Table split into two groups and provided safety-related information by means of structured walkthroughs; each group tackling particular stages of the modelling process. The stages delimited previously were animated via the Safe-DIS workbench so that the members of the Round Table could move through the 'motions' of modelling and enter safety-related information pertinent to each stage as they arrived at it. In the afternoon session the groups 'swapped over' to review and add comments to the information gathered in the morning session.
Some of the information acquired has already been animated as Safety Labels; other information has helped to further refine our structure of the modelling process and to highlight more specific safety-critical aspects and has also prompted ideas for further functions to be incorporated into the Safe-DIS system.
Asides from the knowledge acquisition sessions the Round Table were given demonstrations of the Safe-DIS workbench and its current functionality; this led to discussions of what actual users of simulation software would like to see added by means of 'safety mechanisms' and paved the way for the production of the systems specification document.
We also heard presentations from two members of the Round Table; in December Richard Marshall described a case study of drainage design and in March Jas Mann outlined the basics of Real Time Control and told us about the work of an RTC steering group whose progress we shall watch with interest.
We have developed a system, the Safe-DIS workbench, which at once serves as a research and development environment and as a prototype of the final system. The workbench facilitates the simultaneous acquisition, animation and validation of information. The workbench also acts as an interface to and between several animation mechanisms and the expanding knowledge base. The mechanisms implemented so far are; (i) an expert system performing automated procedures - used to help in approximating 3-D turbulent network with appendages to a 1D stable system that emphasises the backbone of the network; (ii) the introduction of safety labels to the modelling process in line with the principles of explanation and justification - the system interjects with warnings or even intervenes to prevent hazardous use of the simulation software, in a manner sensitive to the experience of the user; (iii) facilities for information retrieval and browsing are developed and will be made accessible from all parts of the system - these will allow the user to navigate the vast quantities of texts available including official guidelines, pertinent legal documents and articles from learned journals, (exploiting the potential granted with the advent of the Internet).
Over time we plan to investigate not only the sophistication, power and number of these modules but also wish to place import on developing interactions between them so as to increase their effectiveness in ensuring safety. Our ongoing research into multi-agent architectures suggests that they may provide a suitable framework for the instantiation of such interactions.
Recent project work has been involved with the production of a system specification document. The details of this specification have been arrived at after initial knowledge acquisition, system prototyping and crucially, demonstrations to, and ensuing discussions with, members of the Round Table.
Although this document describes the core of the Safe-DIS system we believe that there is scope for ongoing development and refinement of the system's particular functions in the light of further knowledge acquisition and consultations with end users.
The next review of the Safe-DIS project is scheduled for 10 April 1995. The next meeting of the Round Table will take place at the University of Surrey on 8 June.
One of the members of the Safe-DIS team will be attending a one-day meeting on 'Intelligent Agents and the Next Information Revolution' on May 9 at Manchester Metropolitan University. The presentations include the role of agents in information systems, co-operating agents, agents and HCI(human computer interface), autonomous agents and information management. This would provide some ideas and techniques for building an agent-based information system to make safety related information available to design engineers in a timely and efficient manner.
We are planning to visit Wallingford Software; (i) to elicit more information regarding the safe use of the simulation package Hydroworks; and (ii) to discuss issues that may arise when the Safe-DIS modules are to be integrated with Hydroworks.
We also hope to make some 'on-site' visits to members of the Round Table so as to gain a better appreciation of their work.
We are carrying out a functionally-orientated investigation into the language used in safety-related communications at the levels of terminology, lexico-grammar and pragmatics.
Our investigations into the language used to communicate safety-related information has led us to hypothesise the existence of a specialist 'Language of Safety' - the subsequent elaboration of which will provide insights into the nature of safety-related information and prompt the development of improved techniques for the elicitation and dissemination of safety-related information.
Multi-Agent Architectures for Ensuring Safety
Simulation of water networks involves complex numerical processing, qualitative reasoning, text analysis and extensive use of heuristics. The user is expected to be fully aware of input values for setting up simulations and should be able to interpret the output of the system. The knowledge for selecting input data and for interpreting output comes from disparate disciplines.
A user of simulation software modelling network needs to execute a number of interdisciplinary tasks. Each of these tasks requires proactive contribution from different knowledge sources and requires an element of co-operation and negotiation among the various knowledge providers. In a multiagent system, each knowledge source whether it be a data set, a program or textual material can be viewed as an agent. In computational terms, an agent may be represented as an autonomous, self-contained, reactive, pro-active computer system.
Currently, we are extending some of our previous work on multiagent systems to create an architecture that will support the social activities of a diverse group of agents. We intend to demonstrate how agents would interact socially with each other by referring a well-known problem in logistics; a number of trucks that operate on a road network to deliver goods. Here is a society of agents that make plans, perform various reasoning operations, and interact with each other.
The long-term goal of this research is to use the architecture for simulating the behaviour of pipe networks, whose operation in real-life may hazard the safety of the users of the network.
By designing and constructing two prototype models, we have been able to carry out the process of knowledge acquisition in conjunction with knowledge animation. In effect we are using this procedure to build a safe system in real time; as the experts offer safety related knowledge it is entered into the system. In this way the experts can update, and correct the knowledge, and also evoke discussions concerning it.
We would like to acknowledge the contributions of the following people to the Safe-DIS project;
Bob Armstrong, Phil Gelder, John Hately, Philip High, Jas Mann, Richard Marshall, Bob Sugden, Steve Griffin, and Paul Holmes-Higgin.
Contact addresses for the Safe-DIS Group
Khurshid Ahmad, Steve Collingham, Andrew Salway, Indran Selvaratnam