Concurrent Bandwidth Feedback for Complex Manual Control Tasks

Abstract

Augmented feedback has been demonstrated to be an effective technique for improving human subject performance for a variety of simple and low-dimensional tasks, but more complex and realistic tasks are rarely explored in the literature. Instead of simply providing additional guidance, augmented feedback alerts operators to critical features of a task that they may not otherwise be aware of. However, past research has revealed a significant caution --- that many forms of augmented feedback lead to the guidance hypothesis, which manifests as decreased performance when the feedback is removed.

The research presented here explores a novel approach to a specific type of augmented feedback, called concurrent bandwidth feedback, and how it might be effectively applied to avoid causing the guidance hypothesis when training for complex tasks. Concurrent bandwidth feedback is provided to an operator in real-time, during task execution, when a specific signal deviates outside of an acceptable range of values. This ``Instructor Model'' of feedback highlights display elements when urgent attention is needed from the operator and, consequently, appears less frequently as operator performance improves. Through the analysis of four human-in-the-loop experiments, we establish that subjects exposed to concurrent bandwidth feedback immediately and significantly improve task performance. By varying the functional task complexity in an aircraft flight task, we demonstrate that the improvements in operator performance increase with task complexity. Investigation of immediate and 24-skill retention shows that concurrent bandwidth feedback does not result in the guidance hypothesis. While increased human performance usually results in higher levels of cognitive demand, subjective workload measurements and objective secondary task performance indicate that our subjects did not experience additional cognitive load when using our novel concurrent bandwidth feedback.

Control theory-based modeling techniques are explored to further understand and predict the effects of concurrent bandwidth feedback. To explain the increased task performance, the Structural Model of the human pilot was extended by introducing a new control block that models the concurrent bandwidth feedback received by the operator. Using data from our aircraft flight task and the Structural Model, we show that exposure to the concurrent bandwidth feedback results in increased error sensing and gain compensation, raising the resultant crossover frequency and ultimately improving pilot performance.

In all of our studies, subjects exposed to concurrent bandwidth feedback had higher levels of performance, did not experience increased workload, and did not suffer from the guidance hypothesis, indicating that concurrent bandwidth feedback can be used as an effective training technique for complex manual control tasks.