• About
  • Advanced Search
  • Browse Proceedings
  • Access Policy
  • Sponsor
  • Contact
  • Adaptive and optimalizing behavior of the human operator in compensatory tracking

    Paper ID



    • D. T. McRuer
    • E. S. Krendel
    • D. Graham


    Systems Technology, Inc.






    In the earliest days of aviation, there were repeated suggestions that because of limitations on the capabilities of the human operator, the stabilization of aircraft should be accomplished automatically.1,2 It soon turned out, however, that the use of a human controller was a practical means for achieving success, and interest in the design of automatic stabilization and control equipment for airplanes declined and remained at a low level for a long time.3 Some say that history tends to repeat itself. When space flight was first contemplated, it was widely thought that the ability of the human operator was so markedly limited, and that the cost of maintaining a suitable environment for him in space was so high, that the participation of man in space flights would not prove to be worthwhile. More recently, there has been renewed interest in the remarkable ability of the human operator to enhance the probability of space mission success and to perform control functions in an adaptive and optimalizing fashion. Adaptive and optimalizing control for maneuvers in which the characteristics of the input or controlled element may vary over a wide range may be a ne cessity. Such maneuvers possibly include launch, lunar landing, and reentry, among others. The several reasons for assigning man â role in space, derived entirely from engineering considerations, were summarized in 1959 in Ref. 4. The authors pointed to the human qualities which could be used to augment system reliability, such as failure detection, replacement, and repair; as well as to the human’s ability to organize data, to make decisions, to observe, and to remember. They also emphasized man’s capability as an adaptive controller and power actuator, but concluded that more research was required in order to provide a definitive basis for design trade-off studies concerning the allocation of functions bètween the man and the machine. We do not propose here to enter into any dispute over manned vs. unmanned space flight, but instead purpose only to report some recent research results on the performance of the human operator as a controller. A mathematical model of the human operator, quantitatively expressing his capabilities and limitations in given tasks would allow the best design of displays, control systems, and dynamic characteristics of the vehicle. A number of investigators have made attempts to find more or less elaborate models. ELKIND5 has recently reviewed the status of these activities as they are known in the United States. Of fundamental importance is a quasi-linear model of human operator response in single-loop compensatory tracking of random appearing, limited bandwidth inputs. Some, but not all, flying tasks partake of the nature of compensatory tracking, i.e., the pilot moves the controls so as to null an “error” which is displayed to him, or which he otherwise perceives without being separately aware of the input and output of the system. Even when the flying is not actually compensatory, performance in compensatory tracking of random-appearing waves may represent a conservative estimate of operator capabilities. (This is.because given information about the input and output, as in a pursuit display, or being able to detect a measure of coherence in the input, such as in connection with a single sine wave, the operator can adopt different forms of response and may achieve a higher level of performance. In effect, he is then able to cancel otherwise inherent response lags and delays.6) ' Simulator and flight tests have repeatedly shown that in single-loop or uncoupled multiloop, error cancelling, manual control systems which do not make excessive demands on the capabilities of the human operator, the organizations and adjustments adopted by the pilot tend to be the ones required by “good” feedback systems.7,8’9,10 A quasi-linear model of the human operator then would allow the powerful methods of feedback system analysis to be used for the prediction of pilot-vehicle control system performance.11’12 In the case of more complicated multiloop systems, direct experimental data on pilot Organization and adjustment do not yet exist. Nevertheless, a comparison of experimental data with inferences based on an extended model strongly suggest that the pilot model and analysis methods have a continuing practical validity.