Probabilistic Analysis of Hazard Situations

J. K. Kuchar (jkkuchar@mit.edu),
R. J. Hansman (rjhans@mit.edu)

Research Objectives

Develop a methodology that can be used to design and evaluate systems that alert flight crews to external hazards. The framework incorporates threats generically in order to provide an evaluation method that can be applied to terrain, traffic, weather, or other aviation hazards. Other applications are also possible (e.g., IVHS or process control)

Approach

A probabilistic analysis methodology has been developed to examine the fundamental relationships between the parameters affecting alerting system design (such as sensor accuracy or alert threshold method). The methodology is created through an investigation of the basic issues relating measurement accuracy, expected aircraft performance, hazard severity and extent, and situation geometry. The methodology can be used to evaluate alerting systems both in terms of achievable false alarm rate and by highlighting areas where further improvements in system design may be of value. Numerical integration or Monte Carlo Simulation can be used to obtain quantitative results.

Probability of Incident Curves

System Operating Characteristic (SOC) Curve

Accomplishments

The figures show example applications of the methodology to situations with the Ground Proximity Warning System (GPWS) and the Traffic Alert and Collision Avoidance System (TCAS). In the GPWS example, probability density functions of the pilot's response to an alert are generated and used to determine the probability that a terrain collision incident will occur. In the TCAS example, a situation is used in which an intruding aircraft is descending toward the own aircraft but may level off safely with some probability. The tradeoff between False Alarms and Correct Detections can be shown using a System Operating Characteristic (SOC) curve. As the SOC curve shows, the updated TCAS Version 6.04A threshold is located such that it produces a high probability of a Correct Detection while minimizing the probability of a False Alarm.

GPWS Situation

Pilot Response Probability Density Functions

GPWS Probability of an Incident

TCAS Encounter Situation

TCAS SOC Curves

Significance

The probabilistic methodology can be used to examine the tradeoffs between improvements in the design of the system. For example, the benefit from increased sensor accuracy or decreased response latency can be quantified and compared. Changes in the alerting logic can also be investigated (quantifying, for example, the effect of including information regarding the intended path of a threat).