The effect of level of automation and adaptive automation on performance in dynamic control environments



Journal Title

Journal ISSN

Volume Title


Texas Tech University


Level of automation (LOA) designates the degree of human and computer interaction in controlling an automated system and has typically been examined in a binary fashion; therefore, either the human or computer is assigned to a given task. Recently, studies have investigated intermediary levels of automation with the intent of keeping both the human operator and computer involved in systems performance to promote operator situation awareness and reductions in out-of-the-loop performance problems.

Allocating automated system (manual) control between human and computer servers for varying durations of time throughout system functioning with the intent of improving human operator performance has been labeled as adaptive automation (AA). The effect of manual control allocations on operator performance during fully automated operations following various scheduling strategies has been also been investigated as a means of promoting operator monitoring performance in working with automated systems.

The objective of this study was to examine the interaction between LOA and AA. An experiment was conducted to investigate the impact of various levels of automation and AA strategies on human/machine system performance, operator situation awareness and workload in dynamic control tasks within a multitask environment. A secondary goal of this study was to assess the relation of primary task performance, situation awareness and workload with secondary task performance.

Thirty subjects performed a dynamic control simulation and secondary gauge monitoring task, simultaneously. Testing involved trials at five levels of automation ("Batch Processing," "Shared Control," "Blended Decision Making," "Supervisory Control" and "Full Automation") allocated during manual functioning for different automation allocation cycle times (AACT) comprising 0%, 20%, 40%, 60% and 100% of total task time, respectively.

Results revealed AACT to have little effect on SA and performance, as compared to LOA, yet it was the driving factor in changes in subjective workload and secondary task performance. Level of automation (as a main effect) had little influence on workload and secondary task performance, yet it accounted for a significant portion of the variance in primary task performance and SA. The combined effect of AACT and LOA on all response measures was not additive in nature. Interestingly, the LOA yielding the "best" overall performance ("Batch Processing") did not do so at the AACT producing superior functioning (100% automation cycle time).



Cognitive learning, Human-computer interaction, Control theory, Automatic control, Learning