Real-time hierarchically distributed processing network interaction simulation

  • Authors:
  • Wayne F. Zimmerman;Chung-I Wu

  • Affiliations:
  • Member of Technical Staff, Automated Systems;Member of Technical Staff, Flight Computer Systems and Technology, Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California

  • Venue:
  • ANSS '88 Proceedings of the 21st annual symposium on Simulation
  • Year:
  • 1988

Quantified Score

Hi-index 0.00

Visualization

Abstract

The Telerobot Testbed is a hierarchically distributed processing system which is linked together through a standard, commercial Ethernet. Standard Ethernet systems are primarily designed to manage non-real-time information transfer. Therefore, collisions on the net (i.e., two or more sources attempting to send data at the same time) are managed by randomly rescheduling one of the sources to retransmit at a later time interval. Although acceptable for transmitting noncritical data such as mail, this particular feature is unacceptable for real-time hierarchical command and control systems such as the Telerobot. Data transfer and scheduling simulations, such as token ring, offer solutions to collision management, but do not appropriately characterize real-time data transfer/interactions for robotic systems. Therefore, models like these do not provide a viable simulation environment for understanding real-time network loading. A real-time network loading model is being developed which allows processor-to-processor interactions to be simulated, collisions (and respective probabilities) to be logged, collision-prone areas to be identified, and network control variable adjustments to be reentered as a means of examining and reducing collision-prone regimes that occur in the process of simulating a complete task sequence. The phase-one and phase-two development results are presented in this paper. Results include 1) the theoretical foundation for the network flow model, 2) an overview of the simulation design and constraints, 3) the software design, and 4) a summary and analysis of actual results. Ultimately, the simulation will be used to examine potential loading problems as out-year demo performance improvements cause increased data traffic. The simulation will also provide a systematic means of managing resulting loading problems.