Search

The Cyberlink is an alternative controller that employs electromyographic, electrooculographic and electroenceptialo- graphic signals measured at the operator's forehead. The Phase I SBIR developed a pre-production prototype and demonstrated that operators are able to use it for continuous (point) and discrete (click) computer control with promising efficiency and accuracy. The Phase II effort transitions this technology to a mature system that is available commercially. A dry electrode/headband system was developed for signal acquisition. A low noise bioamplifier and micro-controller performs signal conditioning and initial processing. Windows 3.1 and 95 drivers were developed to permit the Cyberlink to interface with third-party software. A suite of calibration, training and adjustment utilities also operate under Windows. The final evaluation compared the Cyberlink and a standard mouse in a two-axis target acquisition task that employed a Fitts' Law paradigm. Both mouse and Cyberlink performance could be modeled with Fitts' Law, although the mouse was about four times faster. This is an encouraging result, given the subjects much greater experience with a mouse. The Cyberlink is currently being employed as an assertive and therapeutic technology. Its rapid discrete input capability is being explored for several Air Force applications that would benefit from hands-free control.

Two experiments were performed on separate moving-base flight simulators at Wright-Patterson AFB to determine the effects on air to air tracking tasks of motion-reducing drive logic (washout filters) in the roll and sway degrees of freedom. Measurements included not only error and control performance, but also pilot behavior (describing functions and fitted parameters) and pilot subjective evaluations. Excellent consistency of data resulted across the two facilities, and between very thoroughly practiced non- pilots and experienced military pilots. The results indicate that rotary motion cues are used primarily in the role of stability augmentors (i.e., as rata dampers) and that lateral specific-force cues below 0.1 Gy are ignored or have small effects. Limited data indicate that grossly spurious motion distortion due to washout filter dynamics were rated and performed worse than no motion at all. Optimum combinations of attenuation and first-order washout filtering were found for the roll motion drive logic. An adaptive nonlinear logic was developed and validated for the sway drive logic.

An experimental and analytical study was undertaken to test a model for the pilot's use of motion cues in roll-axis tracking tasks. Simulated target-following and disturbance-regulations tasks were explored with subjects using visual-only and combined visual and motion cues. The effects of motion cues on task performance and pilot response behavior were appreciably different for the two task configurations and were consistent with data reported in earlier studies for similar task configurations. The 'optimal-control' model for pilot/vehicle systems provided a task-independent framework for accounting for the pilot's use of motion cues. Results were consistent with the hypothesis of attention-sharing between visual and motion variables.

An experiment is presented in which effects of roll motions on human operator performance were investigated. The motion cues considered were the result of commanded vehicle motion and vehicle disturbances. An optimal control pilot-vehicle model was used in the design of the experiment and to predict system performance before executing the experiment. The model predictions and experimental results are compared. Seventy-eight percent of the model predictions are compared. The high correlation between model predictions and system performance indicate the usefulness of the predictive model for experimental design and for prediction of pilot performance influenced by motion cues. (Author), Presented at the Annual Conference on Manual Control (13th) 15-17 Jun 77, Cambridge, MS.

Presented at the IEEE Conference on Decision and Control, New Orleans, LA. 7-9 Dec 77.

By applying a maximum, likelihood approach to identification with empirical data from a tracking task, the output signal uncorrelated with the input signal (a definition of human operator remnant) can be determined. To obtain this remnant signal, a linear, stationary model describing the human is utilized. The innovations signal is computed from the difference in the empirical data and the model's output. The remnant can then be identified using the innovations sequence by computing the component of the output signal which is uncorrelated (or orthogonal) to the input signal. Data from a Roll Axis Tracking Simulator is analyzed and remnant is identified to two phases of tracking (with and without motion information). (Author), Presented at the IEEE Conference on Decision, 10-12 Dec 75, Houston, Tex.

The optimal control model for pilot/vehicle systems has been modified to allow a different treatment of motor related sources of pilot randomness. Specifically, the concept of 'pseudo motor noise' has been implemented to provide a model parameter related more directly to uncertainties about the pilot's control input; this noise is treated as a disturbance on control rate. The revised model is able to provide a good match to low frequency 'phase droop' and to other measures of system performance and pilot response for previously explored tracking tasks. Data obtained in an AMRL experimental study of roll axis tracking are analyzed. The effects of motion cues on tracking performance are modeled by (1) an increment of 0.05 seconds to the pilot time delay, and (2) inclusion of sensory variables that are likely to be provided by motion cues (position, rate, and acceleration of the vehicle roll angle in these experiments).

The objective of this work was to develop a systems engineering based methodology for quantifying and modeling effects of cognitive loading on human electrocortical responses in systems engineering terms. The visual-cortical response was selected as the input-output channel around which the methodology was developed. A steady-state input ( a continuous sum of 10 sine waves) and a transient input (a train of pulses) were used for system stimulation. The human occipital EEG (surface electrodes at Oz and mastoids) was used as the cortical describing functions (gain and phase) and remnant spectra (background EEG). A stimulus parametric investigation indicated that sensitivity to stimulus modulation depth is unequal across frequencies, and that relatively low stimulus intensity and depth of modulation are desirable to reduce response saturation. Comparisons between transient and steady-state simulation revealed that both forms produced functionally related responses, suggesting that the visual-cortical response contains a measurable linear portion. Time domain amplitude changes corresponded to transient and steady-state gain changes. To investigate cognitive loading effects three tasks were utilized: manual tracking, grammatical reasoning, and supervisory control. Changes in visual-cortical response measures related to task loading were observed. With increased cognitive loading, alpha responders generally showed reductions in alpha band gain and remnant, and increases in beta and gain. Non-alpha responders showed increases in beta band gain only., Prepared in collaboration with Bolt Beranek and Newman and Idaho Univ.

A laboratory tracking study was accomplished to determine the degree of correlation between mean square tracking error and a hit scoring metric. Prior studies had established high correlation between the hit scoring metric and miss distance in a simulation of antiaircraft artillery (AAA) systems. Further, in this same study mean square tracking error and the hit score metric was uncorrelated. The present study used a Gaussian forcing function and twenty subjects in a laboratory tracking task. A prediction of hit scores based upon mean square tracking error was developed based upon the assumption of Gaussian tracking behavior by the trackers. The predictions of the theory and the results of the tracking studies compared favorably. Thus, if tracking error is Gaussian stationary the hit score can be predicted from the mean square tracking error. If the hit score and mean square tracking error are uncorrelated, then the tracking behavior of the subjects is not stationary Gaussian. (Modified author abstract)