Preliminaries - Real-Time Rendering: Computer Graphics with Control Engineering - page 17

system can be approximated by a linearised model with corresponding conceptual
software representations of actuators and sensors. Although the feedback control
architecture was provided for generic software systems, the entire work focuses on
web server applications.
Abdelwahed et al. [3] proposed a generic online control framework to design
self-managing computer systems. The control actions governing system operations
were obtained by optimising system behaviour as forecasted by a mathematical
model over a specified time horizon. The case studies cited deal with power manage-
ment under time-varying workloads and signal detection accuracy and latency levels.
Since computer systems in networked environments are gaining importance due to
increasing Internet usage, Li and Nahrstedt [4] proposed a task control model to illus-
trate the dynamics of QoS adaptations using digital control theory. The objective was
to provide optimum resource allocation to tasks in a distributed environment where
multiple applications compete for and share limited system resources, thus ensuring the
best user experience and efficiency. A proportional, integral, derivative (PID) control-
ler was used to achieve the desired performance objectives relating to the QoS metrics.
Hellerstein et al. [5] provided a comprehensive overview of the challenges in con-
trol engineering of computer systems. Similar studies were reported by Abdelzaher
et al. [7], Lu et al. [8], and Karamanolis et al. [9].
2.3.3 c ontRol P RinciPles in c omPuteR g RaPhics s oftwaRe
In Li and Shen's work [10], a fuzzy logic controller serves as an automatic mecha-
nism for controlling error tolerance in hierarchical volume rendering. Volume ren-
dering is a technique for directly displaying a sampled 3D scalar field without first
fitting geometric primitives to the 3D discrete sampled date set. The performance
criterion is a user-defined frame rate that the control system will strive to achieve
based on adjusting the error tolerance.
Sort-last rendering is a computer graphics applications technique for rendering
extremely large datasets in clusters of computers, usually in a distributed environ-
ment. Kirihata et al. [11] showed that it is possible to use feedback control to harness
large data transfer processes in sort-last rendering.
Another example of the adoption of control principles in computer graphics
software is the work by Dayal et al. [97]. They proposed an adaptive form of
frameless rendering with the potential to increase rendering speed dramatically over
conventional interactive rendering. This is done without the rigid sampling patterns
of framed renderers and by allowing sampling and reconstruction to adapt with very
fine granularity over spatial-temporal colour changes. A sampler uses closed-loop
feedback to guide sampling toward edges or motion in the image to maximise ren-
dering efficiency.
To date, little research has focused on the adoption of control principles in com-
puter graphics applications related to rendering. While the potential benefits are
immense based on a broader perspective in which control techniques have been used
successfully in generic software, the challenges usually lie in specific applications
that require in-depth understanding and appropriate modelling before the control
concepts may be introduced.