Markov analysis of human-in-the-loop system performance

Pilot interaction with complex vehicles involves information perception and understanding, as well as decision making to select and execute the desired action. These decisions and actions are often time-critical and require an accurate response. When designing a complex system, the analysis of human-in-the-loop system performance is important during early-stage system design to assess the impact of varying levels of automation, redundancy, and task allocation. We have integrated several human performance models with a model of a piloted vehicle to analyze human-in-the-loop performance using Draper Laboratory´s Performance and Reliability Analysis via Dynamic Modeling (PARADyM) toolkit. This approach provides a framework for understanding the effects of a vehicle component failure or human error as it propagates through a complex system. Vehicle and human performance models, which include a model of the Space Shuttle Orbiter lateral flight dynamics, visual and vestibular perception, rule-based judgment and decision making, and pilot action, were implemented using MATLAB/Simulink?. Trajectory scenarios were simulated for analysis with and without instrumentation failures, and with and without human errors. The resulting pilot-vehicle performance during scenarios with a component failure was compared to a baseline (no failure) trajectory. Performance thresholds were specified to determine whether the resulting vehicle trajectory represented degraded performance that was within the specified bounds (operational) or outside the bounds (resulting in system loss). At the present stage, this analysis methodology is viable as an early-stage design tool. However, if associated with experimentally validated models for both the human performance and vehicle dynamics, this approach has the potential for a mission and configuration design analysis tool.