AN INTELLIGENT KNOWLEDGE BASED SYSTEM FOR MARINE COLLISION AVOIDANCE

A prototype expert system for collision avoidance has been developed, incorporating the International Rules for Preventing Collisions at Sea. This system simulated an encounter between two vessels whose control/response characteristics were read into the system. A built-in rule structure evaluated the situation independently for each vessel at each time-step, advising the simulator module that then calculated the relevant parameters and updated the knowledge base for each vessel; each vessel had access only to such information about the other as would be available through ships' instrumentation. A multi-window screen display was used to show: tracks of the two vessels; current speed, direction and status information; options available; and a review of progress of the encounter, if requested, indicating reasons for decisions made. A theoretical evaluation, based on a sample of 500 randomly generated encounters, has shown this prototype to be highly efficient at handling most encounter situations; it has also identified areas for improvement of the decision-making logic. A second prototype, incorporating various changes, differs in one major respect from the first: the rule base has been separated out from the program logic, leaving a content-free inference engine. Each rule is now defined explicitly in terms of: description of current (new) status; antecedents; consequent action (generally a sequence, involving other rules). Planned extensions to the user interface will allow modification and extension of the rule base by competent persons. The expert system has been written in the 'C' programming language, on a microcomputer system capable of operating in real-time on-board ship. The prime objective is to provide an expert advisor for mariners, capable of acting on sensory input from all available channels, including information provided by an operator via a keyboard, and advising on the preferred course of action.

• Supplemental Notes:
  • Conference paper
• Authors:
  • Blackwell, G K
  • Colley, B A
  • Stockel, C T
• Conference:
  • International Conference on Maritime Communications and Control
  • Location: London, England
  • Date: 1988-10-26 to 1988-10-28
• Publication Date: 0

Media Info

• Features: References;
• Pagination: 8p.+discuss.

Subject/Index Terms

• TRT Terms: Crash avoidance systems; Expert systems; Traffic regulations
• Uncontrolled Terms: Rule of the road
• Subject Areas: Law; Marine Transportation;

Filing Info

• Accession Number: 00659380
• Record Type: Publication
• Source Agency: Maritime Technical Information Facility
• Files: TRIS
• Created Date: Jul 21 1994 12:00AM