|
|
 |
|
|
|
Our team is developing a human performance modeling tool, the Human Oculomotor Performance (HOP) model, to accurately predict human performance in complex multi-tasking and divided attention environments with several displays and long sequences of information acquisition activities and crew responses.
Our unique approach to the development of such a model is building up the requisite model capabilities by linking model development closely to analyses of actual operator performance while managing faults in a part-task simulation of a spacecraft ascent. Using eye-movement recording, we analyze operator's fixation sequences to recover and characterize the sequence of information acquisition activities that they engaged in to perform the task. We then attempt to replicate these information acquisition sequences with the HOP model, which decomposes the task into the minimal chain of information acquisition activities (fixations on display regions of interest) necessary to complete task subelements, generates actual sequences of fixations to complete that sub elements stochastically (see below), and then stochastically samples from empirically-derived gamma distributions to select the duration of each individual fixation. The sum of the fixation durations for these fixation sequences gives the total time to complete the task.
To generate the stochastic fixation sequences, we set the probability that the model will re-fixate on the current element of interest, regress to the previous fixation, interrupt to service a concurrent task, or proceed to the next fixation in the minimum fixation chain necessary to complete the task. By systematically manipulating these parameters, we have recently been able to simulate (reproduce) the full range of human operator completion times for a fault management task during simulated spacecraft ascents during which the operator had to diagnose the root-cause of a Caution and Warning system event, and then complete the appropriate list of fault isolation and recovery procedures.
|
|
|
