Aspects of Adaptive Automation Support of Air Traffic Controllers David B. Kaber, Ph.D. Department of Industrial Engineering NC State University NASA Langley Core Competency Directors' Visit, September 4, 2003 • • Introduction Adaptive automation (AA) - Dynamic allocation of machine system control to human operator or computer over time with purpose of optimizing performance (Kaber & Riley, 1999). Example system: Ground Collision Avoidance System in fighter aircraft (e.g., F-16) • System monitors and predicts altitude. Provides warnings to pilot. Takes control of flight path. Returns control to human pilot. Historically implemented using binary approach (Parasuraman et al., 1993; Hilburn et al., 1993): Full Automation Automation Manual Control Binary Approach Full Full Full Automation Automation Automation Manual Manual Manual Manual Control Control Control Control Time Why Study AA? • Potential solution to problems with high-level, static automation: • High monitoring workload (Wiener, 1988). Operator complacency and vigilance decrements (Parasuraman et al., 1993). Loss of situation awareness (SA; Kaber et al., 2001). Skill decay (Parasuraman, 2000; Shiff, 1983). Benefits of AA found in prior research: • Lower operator perceived workload in air traffic control (ATC) simulation compared to static automation (Hilburn et al., 1997). Questions remain as to how to effectively implement AA: What functions should be automated? Who invokes automation? What is basis for function allocations (operator workload, SA, etc.)? (Wickens & Hollands, 2000) Improvements in monitoring performance in multiple task scenario (using MAT Battery) compared to static automation (Parasuraman et al., 1993; Hilburn et al., 1993). NASA Langley Core Competency Directors' Visit, September 4, 2003 Grants and Projects Completed • FY01: • FY02: • “Human Response to Adaptive Automation of Information Acquisition Functions and Later Stages of Information Processing” NAG-1-0139; $39,971 (PM: L.J. Prinzel) “Authority in Adaptive Automation Applied to Various Stages of Human-Machine System Information Processing” NAG-1-02056; $50,067 (PM: L.J. Prinzel) NASA GSRP Grant: “Comparison of Physiological and Secondary Task Measures for Triggering Adaptive Automation” NGT-1-01004; $66,000 (TPOC: L.J. Prinzel) Period of performance: 5/15/01-5/14/04. NASA Langley Core Competency Directors' Visit, September 4, 2003 • AA of ATC Information Processing Functions Need - Compare • Approach: effectiveness of AA applied to Developed PC-based simulation of TRACON and secondary various air traffic controller monitoring task. information processing functions, including: 1. 2. 3. 4. Information acquisition gathering of aircraft flight parameter data. (Auto was like radar tracking system.) Information analysis summarizing data, including conflict identification. (Auto was like futuristic EDD or TPA.) Decision making - sorting aircraft in terms of priority for clearance based on potential conflicts. (Auto was like CAA.) Action implementation automated clearances after communication link established (auto was like data link). Keyhole Automation Aid ATC functions were adaptively automated based on monitoring task performance (objective measure of workload). Findings: May have been due to complexity of automation aid and visual attention demands of display. Performance During Manual Minutes 60% 50% Manual Action Decision 10% Analysis 20% Acquisition 30% Manual 40% Action 70% Decision Workload measure very sensitive to auto state changes when AA applied to action implementation. Analysis Performance “best” when AA applied to act of issuing clearances (action implementation). Manual control periods as part of AA of action implementation better than all other conditions and completely manual control. AA of information acquisition significantly reduced controller workload. AA of information analysis function produced highest workload. Acquisition Average Percent Cleared • Workload-Matched AA of Controller Functions 0% Trial 1 Trial 2 Level of Automation Information Acquisition Information Analysis Action Implementation Manual Decision Making FY01 Outcomes • Conference proceeding paper: • Clamann, M. P., Wright, M. C. & Kaber, D. B. (2002). Comparison of performance effects of adaptive automation applied to various stages of human-machine system information processing. In Proceedings of the 46th Annual Meeting of the Human Factors and Ergonomics Society (pp. 342-346). Santa Monica, CA: Human Factors and Ergonomics Society. NASA Technical Publication: Kaber, D. B., Prinzel, L. J., Wright, M. C. & Clamann, M. P. (2002). Workload-matched adaptive automation support of air traffic controller information processing stages (Tech. Pub.: NASA/TP-2002-211932). Washington, DC: NASA. NASA Langley Core Competency Directors' Visit, September 4, 2003 FY01 Outcomes • Book chapter: • Kaber, D. B. & Wright, M. C. (in press). Adaptive automation of stages of information processing and the interplay with operator functional states. To appear in. G. R. J. Hockey, A. W. K. Gaillard & O. Burov (Eds.), NATO Advanced Research Workshop - Operator Functional State: The Assessment and Prediction of Human Performance Degradation in Complex Tasks. Amsterdam: IOS Press, NATO Science Series. Journal article: Kaber, D. B., Wright, M. C. & Clamann, M. P. (in revision). Adaptive automation of information processing functions and operator stress. Submitted to Human Factors (5/31/02). NASA Langley Core Competency Directors' Visit, September 4, 2003 Authority in AA of ATC Functions • Need - Assess human performance and workload effects of various forms of authority in AA of ATC information processing functions. • Automation Feedback Display Computer mandates computer has complete control of dynamic function allocations in ATC task. Human input is irrelevant. Computer suggestion computer initiates function allocations but needs human approval to invoke. Approach: Implemented same forms of ATC function automation as studied in FY01 (info acquisition, info analysis, decision making, action implementation). Developed new version of ATC simulation including feedback display on locus of control and automation. All computer mandates and suggestions for automation based on operator secondary task performance (workload). Findings: Automation of data gathering function (information acquisition) yielded best performance. Automation of sorting of aircraft for clearance (decision making) produced worst performance. Performance hindered by complexity of automation. High visual display demand. Computer suggestions of automation better than mandates, in general. However, when performing task manually, mandates of auto better than suggestions. No differences in workload among types of automation of forms of authority. Average Percent Aircraft Cleared • 100% 90% 80% 70% 60% Auto AA 50% Manual AA 40% 30% 20% 10% 0% Information Acquisition Information Analysis Decision Making Action Implementation Autom ation Type 100% Average Percent Aircraft Cleared Invocation Authority in AA 90% 80% 70% 60% Auto AA 50% Manual AA 40% 30% 20% 10% 0% Mandate Suggest Authority FY02 Outcomes • Masters thesis: • Technical Report: • Clamann, M. P. (2002). “The Effects of Intermediate Levels of Invocation Authority on Adaptive Automation of Various Stages of Information Processing.” Kaber, D. B. & Clamann, M. P. (March 2003). Authority in adaptive automation applied to various stages of humanmachine system information processing. (Final Rep.: NASA Langley Research Center Grant #NAG-1-02056). Hampton, VA: NASA Langley Research Center. Conference proceeding paper: Clamann, M. P. & Kaber, D. B. (in press). Authority in adaptive automation applied to various stages of humanmachine system information processing. To appear in Proceedings of the 47th Annual Meeting of the Human Factors and Ergonomics Society. Santa Monica, CA: Human Factors and Ergonomics Society. NASA Langley Core Competency Directors' Visit, September 4, 2003 Current Research: SA and AA • FY03: • “A Situation Awareness-Based Approach to Adaptive Automation” NAG-1-03022; $100,000 (PM: L.J. Prinzel) Subcontract to SA Technologies, Inc. for consultation on development of real-time measure of SA ($9,984). Needs: Define measure of SA in ATC sensitive to dynamic function allocations as part of AA. Empirically assess utility of controller SA measure for classifying forms of AA of info processing functions. Describe impact of AA of ATC information processing functions on controller SA. NASA Langley Core Competency Directors' Visit, September 4, 2003 SA and AA • Approach: Developed enhanced version of TRACON simulation based on input from PM. Studied task analyses on en route control and TRACON (e.g., Endsley & Rodgers, 1994; Endsley & Jones, 1995). Developed query-based measure of controller SA (like Situation Awareness Global Assessment Technique (Endsley, 1995)). Currently conducting experiment: Eight trained subjects working over 3 week period. Experience all forms of AA of ATC task. SA quizzes posed during experimental trials: • Simulation frozen at random points in time. Subjects respond to questions on state of aircraft and environment. Responses compared with actual state of task to evaluate to accuracy of SA. Hypotheses: Lower levels of automation, including information acquisition and action implementation expected to support controller SA. Manual control periods as part of AA expected to increase controller SA. Information analysis automation may degrade system state comprehension. Decision making auto may degrade operator projection of future system states. Enhanced ATC Simulation Data box (flight parameters). Trajectory Projection Aid. Radar scan line. Control box for aircraft clearances. Command history and feedback. Automation aid & status display. Aircraft being cleared. NASA Langley Core Competency Directors' Visit, September 4, 2003 FY03 Outcomes • Masters thesis: • McClernon, C. K. (in preparation). “Situation Awareness Effects of Adaptive Automation of Various Air Traffic Control Information Processing Functions” Journal Article: Kaber, D. B. & Endsley, M. R. (in press). The effects of level of automation and adaptive automation on human performance, situation awareness and workload in a dynamic control task. Submitted to Theoretical Issues in Ergonomics Science (12/15/01). NASA Langley Core Competency Directors' Visit, September 4, 2003 Future Research • • • • Develop real-time probe measure of controller SA for use in research on AA. Conduct experiment involving real-time assessment of controller SA as basis for triggering dynamic function allocations (manual and automated control allocations). Compare results of workload-matched AA and SAmatched AA support on controller information processing. Assess performance, workload and SA effects of AA of multiple ATC information processing functions, simultaneously. NASA Langley Core Competency Directors' Visit, September 4, 2003 Contact and Web Site Information • David B. Kaber, Ph.D. Associate Professor Department of Industrial Engineering North Carolina State University 2401 Stinson Dr. 328 Riddick Labs Box 7906 Raleigh, NC 27695-7906 • Faculty Web Page: http://www.ie.ncsu.edu/kaber/ • FY01 Results Page: http://people.engr.ncsu.edu/dbk aber/AA/ Tel.: (919) 515-2362 FAX: (919) 515-5281 e-mail: dbkaber@eos.ncsu.edu NASA Langley Core Competency Directors' Visit, September 4, 2003