Abstract:
|
Standard discrete-time control laws consider periodic
execution of control jobs. Although this assumption
simplifies the control design and the resource utilization analysis
for later implementation, it leads to a conservative usage of
computing resources. On the contrary, event-driven control
offers controllers with a tighter resource utilization. However,
job executions are no longer periodic, and predicting their
computing requirements is crucial for efficient implementation
in severely limited computing systems.
Sampling intervals for event-driven control systems show
different patterns, ranging from chaotic behaviors to periodic
oscillatory patterns, named equilibrium sampling interval sequences
(ESIS). Focusing on resource demands predictability, in
this paper we identify the conditions for event-driven controllers
to exhibit ESIS, and provide methods to characterize and
compute them. Finally, we study the transitions from ESIS to
chaotic sampling. Simulated experiments illustrate the paper
contributions. |