|
What to Do What It's About What To Notice What Works Best |
Press either button to do this tracking task. Try to keep the cursor (lower line) under the moving target (upper line). At some point during the tracking the cursor will disappear, at which point you will have to control the cursor "open loop" (without any "feedback" regarding the cursor's actual position). In the "Proportional" task the position of the cursor is proportional to the position of the mouse (indicated by the arrow on screen); in the "Integral" task the position of the cursor is the time integral of the position of the mouse.
Tracking in the Integral control condition is much more difficult than tracking in the Proportional control condition. It may take several practice sessions in the Integral control condition until you are able to control the cursor with a reasonable level of skill; the best I have been able to do is get the root mean square (RMS) error in the closed loop phase of the Integral control condition down to about 25, meaning that the cursor was, on the average, no more than about 25 pixels from the target.
A graph of the data (target, cursor and mouse position varying over time) appears at the end of each tracking session (which lasts approximately 1 minute). The vertical line indicates the point at which the cursor disappeared. Measures of control (in terms of RMS deviation of cursor from target) for the period before and after disappearance of the cursor are shown below the graph.
This demo shows that "open loop" control is an illusion. Open-loop control appears to occur in situations (such as the Proportional condition, where control -- measured as RMS deviation -- is about the same before and after the cursor disappears) where it is possible to control a perceptible correlate of the cursor (the visible mouse arrow). When no correlate of the cursor can be perceived (as in the Integral condition) there is no control of the cursor. Behavior is, indeed, the control of perception.
Notice that you can achieve good control in the "open loop" phase of the Proportional control condition becuase you are actually controlling a perception (of the mouse arrow) that corresponds to the position of the cursor. Notice that, in the Integral control condition, you cannot achieve good control even though you are still able to see the mouse arrow in the "open loop" phase because the mouse arrow no longer corresponds to the position of the invisible cursor. The point is that you cannot control "open loop". You cannot accurately generate outputs (mouse movements) that will produce the perception you intend to have (of cursor on target).
This demonstration works best if you take time to practice doing "Integral" control. You should try to convince yourself that, no matter how good you get at Integral control, it is impossible control the cursor "open loop".