This brief paper describes the use of “agents” in supervisory control and data acquisition (SCADA) systems. It is stated that the main motivating factors for using these agents are modularity, autonomy, and proactivity.

SCADA systems in the past were proprietary hardware and software systems sold by a variety of process control companies. With the rise of cheaper microcomputers and industrial control interface standards, the move has been to systems using the Microsoft Windows operating system, and Ethernets with proprietary software and less specialized hardware. The differences in functionality of these systems are minimal, and could be thought of as similar to the differences between word-processing programs. Users typically select predefined function blocks and configure them to handle control tasks. User-created programs can be embedded or can interface to the function blocks. These function blocks can be linked to allow more sophisticated control schemes. Custom user interfaces can be created using a provided graphics configuration program; standard displays are also available with minimal configuration needed. The control system often consists of a network of controllers that can be spread over 100 miles or more, sometimes with sporadic connections.

The paper describes how TRLabs has created an environment for running these agents, and how they interact with the system and each other. I am skeptical of their claims regarding advantages: integration, reliability, versatility, and remote access. In my experience, these features already exist with the software in existing SCADA packages--or in combination with add-on software packages--without the complexity of an agent system.