APPLICATION DESIGN ISSUES - Hard Real-Time Computing Systems: Predictable Scheduling Algorithms and Applications - page 401

The robot workstation described above as an example of open-loop control system
can also be a feedback control system if we close a loop with the camera and use the
current visual data to update the robot trajectory online. For instance, visual feedback
becomes necessary if the robot has to grasp a moving object whose trajectory is not
known a priori.

In feedback control systems, the use of real-time computing is essential for guarantee-
ing a predictable behavior; in fact, the stability of these systems depends not only on
the correctness of the control algorithms but also on the timing constraints imposed
on the feedback loops. In general, when the actions of a system strictly depend on
actual sensory data, wrong or late sensor readings may cause wrong or late actions
on the environment, which may have negative effects on the whole system. In some
case, the consequences of a late action can even be catastrophic. For example, in cer-
tain environmental conditions, under autopilot control, reading the altimeter too late
could cause the aircraft to stall in a critical flight configuration that could prevent re-
covery. In delicate robot assembling operations, missing deadlines on force readings
could cause the manipulator to exert too much force on the environment, generating
an unstable behavior.

These examples show that, when developing critical real-time applications, the fol-
lowing issues should be considered in detail, in addition to the classical design issues:

1. Structuring the application in a number of concurrent tasks, related to the activi-
ties to be performed;

2. assigning the proper timing constraints to tasks; and

3. using a predictable operating environment able to guarantee that those timing
constraints can be satisfied.

These and other issues are discussed in the following sections.

11.2 TIME CONSTRAINTS DEFINITION

When we say that a system reacts in real time within a particular environment, we
mean that its response to any event in that environment has to be effective, according
to some control strategy, while the event is occurring. This means that, in order to be
effective, a control task must produce its results within a specific deadline, which is
defined based on the characteristics of the environment and the system itself.