Model-Based Control - Real-Time Rendering: Computer Graphics with Control Engineering - page 67

5 Model-Based Control
5.1 INTRODUCTION
Research in real-time computer graphics focuses on trading speed of rendering for
image quality but does not address the problem of frame rate stability—a critical
component of the user experience. Common techniques offer “best effort” solutions
to achieve interactive frame rates without any performance guarantee. Consequently,
the onus of finding an optimal solution is left to the application developer if not
totally forsaken. In the absence of a feasible solution, investments in many inter-
active applications such as those from the training, visualisation, and simulation
domains may not yield adequate results. Previous research [48,49] has shown the
importance of maintaining interactive frames in these applications.
Control engineering is a mature field of study with myriad applications in vari-
ous systems that affect our daily lives. Its efficiency when applied to electrical and
mechanical systems in fields as varied as aerospace, defence, communications, and
manufacturing equipment has been proven in numerous industries [50,51] around the
world. Little research relates the adaptation of control theory to real-time computer
graphics rendering. However, in recent years, we noted observable momentum of
cross-disciplinary research in control theory and computer systems [2,7,52,54].
In this chapter, we introduce the concepts of control theory and demonstrate rel-
evant techniques as mechanisms for achieving sustainable performance in real-time
computer graphics rendering.
5.2 CONTROL SYSTEM PERSPECTIVE OF
COMPUTER GRAPHICS RENDERING PROCESS
We consider the computing environment for real-time rendering to consist of a
homogeneous infrastructure consisting of both hardware and software. A simple
representation of the rendering system is shown in Figure 5.1.
The rendering process is modelled as the plant in control taxonomy. Its basic func-
tionality is to process a stream of inputs such as 3D geometry and other rendering data
to create a series of images in real time. Other processes running in the same comput-
ing environment may periodically share the memory and CPU time thereby creating
interruptions that may be represented as disturbances . Furthermore, the rendering
system can receive input from user interaction with the computing environment.
To meet the goal of consistent and sustainable frame rates from the output of the
plant, we introduce a controller as an extension of the system shown in Figure 5.1.
63