CC BY
15
FEATURES OF HUMAN-OPERATOR ACTIVITY WHEN PERFORMING TASKS OF STEP-BY-STEP TUNING OF PARAMETERS OF COMPLEX SYSTEMS Guchuk V.V. (Russian Federation) Email: Guchuk441@scientifictext.ru
Guchuk Vladimir Vsevolodovich - PhD in Engineering, Senior Researcher, LABORATORY OF DYNAMIC INFORMATION-CONTROL SYSTEMS, V.A. TRAPEZNIKOVINSTITUTE OF CONTROL SCIENCES OF RUSSIAN ACADEMY OF SCIENCES, MOSCOW
Abstract: the results of experiments on the study of the processes of step-by-step tuning of parameters by a human-operator are given. It is shown that operators used two tuning tactics to search for global extremum. The first is aimed at finding out the global features of a custom system. To do this, the operator sets several parameter values that are evenly distributed over the entire range. The second tactic is the motion by parameter that tracks the local features of the Quality Score. It is shown that the first tactic is more effective. The features of the parameters adjustment processes are described.
Keywords: the adjustment process, the indicator of quality settings, tactics settings, effects.
ОСОБЕННОСТИ ДЕЯТЕЛЬНОСТИ ЧЕЛОВЕКА-ОПЕРАТОРА ПРИ ВЫПОЛНЕНИИ ЗАДАЧ ПОШАГОВОЙ НАСТРОЙКИ ПАРАМЕТРОВ СЛОЖНЫХ СИСТЕМ Гучук В.В. (Российская Федерация)
Гучук Владимир Всеволодович - кандидат технических наук, старший научный сотрудник, лаборатория динамических информационно-управляющих систем, Институт проблем управления им. В.А. Трапезникова Российская Академия наук, г. Москва
Аннотация: приведены результаты экспериментов по изучению процессов пошаговой настройки параметров человеком-оператором. Показано, что для поиска глобального экстремума операторы использовали две тактики настройки. Первая направлена на выяснение глобальных особенностей настраиваемой системы путем задания нескольких значений параметра, достаточно равномерно расположенных по всему его диапазону, с последующей подстройкой в найденной предпочтительной области значения параметра. Вторая тактика - это движение по параметру, которое отслеживает локальные особенности «показателя качества настройки». Показано, что первая тактика более эффективна. Описаны особенности процессов настройки параметров.
Ключевые слова: процесс настройки, показатель качества настройки, тактика настройки, эффекты.
The work of the human operator in interactive control and monitoring systems is well studied. The literature presents the results of a comprehensive review of its activities, analyzes the effectiveness of user interfaces, discusses the psycho-physiological aspects of interaction with software and hardware [1, 2]. An analysis of specific quantitative indicators of human functioning as one of the links in the control system is also given. The obtained assessments of human capabilities help to create more effective interactive systems, improve the safety of the operation of control systems of complex objects, and contribute to the greater efficiency of the human-operator in such systems [3, 4]. But human activity, not related to the speed of sensorimotor reactions, with the dynamics of problem solving in rapidly changing conditions, was not sufficiently studied, see, for example, [5]. In this
regard, the tasks of setting up complex systems are interesting, the setting quality indicator (SQI) of which functionally depends on the adjustable parameters, and the setting itself takes place in fairly comfortable conditions, i.e. not in a "tight time limit". Here, psychological attitudes, personal labor interests, and features of the organizational behavior of the subjects are more pronounced [1]. Knowledge of the characteristics and features of the operator's behavior in the tuning processes (TP) may allow more successfully developing the technology of human interaction and software and hardware systems for this particular class of tasks [6]. The purpose of the experiments to study the process of tuning parameters by a human-operator was to identify fairly stable features or effects in his actions and assess his potential. The paper presents the results of experiments conducted by the author on the study of TP on the scalar SQI. The quality of the settings was determined by the operator according to the readings of the device. The values of Ci parameters were set by the positions of the potentiometer sliders on the control panel. The functions of SQI Qj were either with one or with two or three maxima. SQI was formed after setting all the parameters and pressing the button on the control panel. The task of the operator was to find the global maximum of SQI or at least its zone. According to the results of a series of experiments, a group of 8 most successful operators was selected. In total, the test group of operators conducted more than 900 experiments, mainly for one- and two-parameter problems. When studying the diagrams for one tunable parameter, it was found that operators used at least two tactics to search for global extremum, which can be called "installation" and "tracking motion". In TP, built on the use of the "installation" tactics, the distribution of samples (Q definition points) does not depend on the local features of the SQI. This tactic is aimed at elucidating the global features of SQI by setting several (4 - 7) parameter values that are fairly evenly distributed throughout its range. First, a zone was selected to search for a global maximum, and then a setting was made. The second tactic is characterized by movement associated with local features of Q. This tactic is aimed at consistently exploring the field of SQI. Examples of the dependence of SQI Q on the values of parameter C are shown in Fig. 1a and 2a. The corresponding TP using the first tactic are shown in fig. 1b and 2b, using the second tactic, are shown in Fig. 1c and 2c. The diagrams of TP are presented in the coordinates "step number" m and "parameter value" C. For the second tactic, the following effects were found. In the zone with positive values of Q increments at each step, tracking of local features is manifested in the tendency to decrease the step length AC (the corresponding zones in Figs. 1c and 2c); 2) in the zone with negative values of Q increments, the opposite effect is observed - a noticeable successive increase in the length of steps (the corresponding zones in Figs. 1c and 2c). We studied the relationship of the first effect with the following factors: the increment of the parameter and indicator at the previous step and their ratio, the length of the path traveled in this zone, the average value of the previous steps, the total (total) and average increment of the indicator, the step number (or the number of steps taken in the zone) . The decisive influence on the length of the next m-th step is exerted by the number of steps taken, as well as in individual cases, the ratio AQm_i/ACm_i at its sufficiently large value. Outwardly adequate TP is a piecewise-power approximation of the dependence AC(m).
Fig. 1. Customization with one SQI maxima
Fig. 2. Customization with two SQI maxima
Tracking the local features of Q manifests itself in a tendency to decrease the step length as long as the AQ value is positive. At the same time, the decrease in step lengths is irregular. This is often associated with a marked increase in the AQ/AC value in the previous step. The most rational TP, built on the use of the "installation" tactic, in which the operator more or less evenly distributed several (4-7) samples over the entire range of parameter values, then selecting a small enough area to search for the global maximum (Fig. 1b and 2b). Of the test group of operators, only two operators regularly used this tactic. Meanwhile, it is most effective. If, on average, 15-18 steps were spent on tuning without using the tactics "installation", then using it would reduce the number of steps to 8-11. When considering the diagrams, a tendency was revealed towards a relative increase in the length of steps in the TP, that is, on average, the step length at the end of the TP is somewhat larger than at the beginning, all other conditions being equal. The effect was tested on a "false" setting. The operator was asked to set up a system for which the effect of changing the parameter on the SQI was reduced so much that some connection of SQI Q with C was still felt, but it was impossible to select sufficiently preferable values (or areas) of the parameter being tuned. Under these conditions, an increase in the length of steps was observed up to 5 times in comparison with the initial steps. The processes of setting parameters are significantly influenced by external factors, conditions for setting up, user interface design features, etc. This should be considered when using the above results. They are qualitative, not quantitative.
References in English / Список литературы на английском языке
1. Krol V.M. Specificity of Switching Attention in Mechanisms of Visual Thinking in Hemispheres of the Human Brain // Human Physiology, 2009. № 4. P. 402-408.
2. Guchuk V. V. Compact visualization of dynamic parameters in monitoring and control systems // Scientific Visualization, 2018. №2. P. 61-69.
3. Guchuk V.V. Applied aspects of the organization of an interactive mode for increasing the security of functioning of control systems of complicated objects // European Science, 2018, № 1. P. 18-21.
4. Dorofeyuk A.A., Guchuk V.V., Desova A.A., Dorofeyuk Yu.A. Assessment of the health of a human operator according to information from the pulse signal of the radial artery / Management in intellectual, ergatic and organizational systems. Taganrog, 2013. P. 173-175.
5. Jofrion A., Dayer J., Fineberg A. Solving optimization problems with many criteria based on man-machine procedures / Issues of analysis and decision-making procedures. M.: Mir, 1976. P. 126-145.
6. Guchuk V.V. Application of algorithms of objectifying expert clustering of Multiparameter objects in the analysis of big arrays of information // Advances in Systems Science and Applications, 2018. № 1. P. 102-109.
References / Список литературы
1. Krol V.M. Specificity of Switching Attention in Mechanisms of Visual Thinking in Hemispheres of the Human Brain // Human Physiology, 2009. №4. P. 402-408.
2. Guchuk V. V. Compact visualization of dynamic parameters in monitoring and control systems // Scientific Visualization, 2018. № 2. P. 61-69.
3. Guchuk V.V. Applied aspects of the organization of an interactive mode for increasing the security of functioning of control systems of complicated objects // European Science, 2018.№ 1. P.18-21.
4. Дорофеюк А.А., Гучук В.В., Десова А.А., Дорофеюк Ю.А. Оценка работоспособности человека-оператора по информации из пульсового сигнала лучевой артерии / Материалы конференции «Управление в интеллектуальных, эргатических и организационных системах». Таганрог: НИИ МВС ЮФУ, 2013. С. 173-175.
5. Джофрион А., Дайер Дж., Файнберг А. Решение задач оптимизации при многих критериях на основе человеко-машинных процедур. В кн.: Вопросы анализа и процедуры принятия решений. М.: Мир, 1976. С. 126-145.
6. Guchuk V.V. Application of algorithms of objectifying expert clustering of Multiparameter objects in the analysis of big arrays of information // Advances in Systems Science and Applications, 2018. № 1. P. 102-109.
