Fundamentals of Manual Control - Fundamentals and Assessment Tools for Occupational Ergonomics - page 286

and grasp it into a mathematical model, one encounters a nontrivial problem, as most of the available
techniques only work out well in the time-invariant case.
The earliest attempts in describing human behavior with control-theoretical models failed to pay
much attention to the adaptivity of human behavior. Research showed that when experiments are not
done under (almost) exactly the same circumstances, the human will adapt and the observed control
behavior will be different. The lack of a systematic approach to this problem resulted in scattered data
and many different and unexpected findings, and theory progressed only slowly.
This changed in the late 1950s and early 1960s when McRuer et al. started working on this problem.
Learning from past experience, they first determined and classified a list of variables that could possibly
have an effect on human behavior. These included environmental variables (e.g., conducting the task in
real flight or in a fixed-base simulator), procedural variables (e.g., subject instruction, practice), and
operator-centered variables (e.g., subject motivation, workload).
Most important to understand the adaptation of the human operator are the task variables, however,
which include:
. The dynamics of the system to be controlled
. The properties (bandwidth) of the forcing function, that is, the signal to be followed (in a following
task) or the disturbance signal acting on the system (in a disturbance task)
. The type of display (e.g., a compensatory display or a pursuit display)
. The type of manipulator
McRuer et al. conducted a massive number of tracking task experiments. In their approach they tried to
very closely control all the variables that could have an effect on human behavior and systematically
varied two of the task variables, that is, the dynamics of the system and the bandwidth of the forcing
function signal. As a result of this approach, human variability decreased significantly and for the first
time insight was gained into how and why humans adapt to changing circumstances. In the following
paragraphs the main results of this research will be elaborated on.
12.4.2 Quasi-Linear Pilot Models
Skill Acquisition The earliest work in this field already showed that a human operator establishes, during
a learning and skill-development phase, a particular control system structure (Krendel and McRuer,
1960). The feedback connections in this system are similar to those, which would be selected for the
development of an automatic controller. The loop closures selected will have the following properties
(McRuer and Jex, 1967):
1. To the extent possible, the feedback loops selected and adjustments made will be such as to allow
wide latitude and variation in pilot characteristics
2. The loop and equalization structure selected will exhibit the highest pilot rating of all practical
loop closure possibilities
3. Delays due to scanning and sampling are minimized
In short, the human will establish, in a learning process, a control system that aims at establishing a
trade-off between the requirements of performance and stability, in the same fashion as a control engin-
eer would design an automatic control system.
Compensatory Tracking Tasks Extensive research has been conducted on the problem of modeling
human control behavior in elementary single-axis compensatory tracking tasks, see Figure 12.10a. In
this task the operator must minimize the difference (error E) between the output (Z) of the system to
be controlled and a reference signal (R). This is the same control situation as described earlier,
Figure 12.7. The double-lined block in Figure 12.10a illustrates that pilot behavior in this closed loop
is essentially nonlinear.