RealTime Computing and Control

To engineers involved in designing, developing, or operating control systems for practical applications, research in control may seem an exercise in mathematical abstractions. As an engineering discipline, however, the connection with the physical world is intrinsic to control. Interfacing with sensors and actuators, implementing advanced algorithms on real-time platforms, dealing with sampling time issues, and other such pragmatic matters may seem to be taken for granted in much of advanced control, but in fact there is an extensive body of research that is concerned with these very topics

All advanced control algorithms today are hosted on digital computing systems, and any discussion of real-time control applications must address the specific issues and challenges associated with digital implementation. The benefits of digital realization are numerous: software-based computing allows more sophisticated control laws; updates and maintenance are rendered easier in many cases; control systems can be made more compact; and control can more readily be integrated with ancillary functions such as information display, system health management, and data recording.

But the digital world brings complications too. For example, the continuous variables of a physical system must now be discretely sampled. Variations in the sampling rate are generally assumed to be negligible, but this is not always the case and a significant adverse impact on control quality can result. Similarly, variable delays arise in the processing of sensory data. In today's processors, even the same sequence of arithmetic operations can take more or less time to execute depending on the state of the processor and operand values.

As control programs become more complex, so do their real-time implementations. Most control system computing platforms are used for several tasks in addition to executing the base control law. Real-time computing and control thus also involves the solution of difficult scheduling problems. Different tasks can have different priorities, desired execution frequencies, and execution times; there may be dependences between them that require interprocess communication and accessing of shared resources; and failures of sensors and actuators and other abnormal situations cannot be allowed to compromise safety.