Ecological Interface Design in Aviation Domains: Improving Pilot Trust in Automated Collision Detection and Avoidance Presenter: Danny Ho Supervisor: Dr.

Ecological Interface Design in Aviation Domains:

Improving Pilot Trust in Automated Collision Detection and Avoidance

Presenter: Danny Ho

Supervisor: Dr. Catherine M. Burns

Advanced Interface Design Lab

Systems Design Engineering

University of Waterloo

© 2003, Danny Ho

Establishing Perspective

On July 1, 2002:Bakshirian Airlines Tupolov 154 collided with a DHL Cargo Boeing 757-200 over Southern Germany

Sequence of events as the two aircraft converged:1) Onboard collision system told Boeing to climb2) Onboard collision system told Tupolov to descend3) Air traffic control (ATC) told Tupolov to climb4) Boeing climbed5) Tupolov climbed6) Collision occurred at 35,000 feet. There were no survivors.

The system FAILED! How?

Outline Defining the Problem and the Objective The Methodology

Ecological Interface Design (EID) Abstraction Hierarchy (AH)

The System Traffic Alerts and Collision Avoidance System (TCAS)

Applying EID to TCAS Approach Implementation EID-Enhanced Displays

Experimental Approach - Proposed Experimental Design Data Analysis

Future Direction & Conclusions

Defining The Problem

Possible Contributing Factors Policies and procedures

North American versus European policy

Human Factors Pilot execution Cognitive performance Interface inadequacies Trust in automation

Defining The Problem

The pilot did not have the necessary information to perform effectively in the automated alerting situation

The pilot didn’t know who to trust TCAS or ATC?

Defining The Objective

To propose display enhancements and evaluate their effects on pilot trust and decision making performance in automated air traffic alerting conditions

It is hypothesized that: An EID-enhanced display will increase decision

making performance and accuracy An EID-enhanced display will increase pilot trust in

the TCAS automated alerts

The Methodology

Ecological Interface Design (EID) A framework for designing interfaces primarily for

complex systems (Vicente, Rasmussen, 1992) Nuclear power plant control (Rasmussen, 1985) Aircraft engineering system (Dinadis & Vicente, 1999) Shipboard command and control (Burns et al., 2000)

Shown to improve operator task performance and conflict detection because it develops a contextual link between information to the trained operator

Uses Abstraction Hierarchy (AH) (Rasmussen 1985) as a design basis

The Methodology

Abstraction Hierarchy (AH) A 5-layered systems approach to component and

interaction representation

WHY?

HOW?

The Methodology

EID: “What data should be extracted, and how should it be presented to help the user understand the system?”

UCD: “How do users perform, and what interface elements can be used to optimize their task performance?”

The EID methodology should create displays that convince pilots to perform a task rather than command them to perform a task

The System - TCAS

Traffic Alerts and Collision Avoidance System TCAS 2 – version 7.0 Internationally adopted and mandated by FAA for all

North American aircraft with capacity exceeding 30 Operates independently of onboard systems/radar

TCAS 1 introduced in 1981 TCAS 1 provides only collision detection TCAS 2 also calculates avoidance maneuver

TCAS - Flight Crew - ATC interaction

TCAS Overview2 Levels of Alerts

-TA : Traffic Advisory

- ‘traffic, traffic’

- RA : Resolution Advisories

- ‘climb’, ‘descend’, etc…

Data Inputs

- intruder range, altitude, bearing

- ownship range, altitude, bearing

Operational Parameters

- protected volume varies with speed

- threat based on time, not distance

- pilot must inform ATC of RA maneuver

- pilot must return to ATC course after RA

- no RA’s under 1000 ft altitude

- system accounts for slow convergences

- if intruder doesn’t react to their RA, ownship RA can be recalculated

TCAS Aural and Visual Alerts

TCAS Display

TCAS in MS Flight Simulator 2002

Applying EID to TCAS

This study introduces a novel approach to applying EID to collision detection and avoidance, dividing the problem into 3 entities

(A)ircraft, (T)CAS, and (E)nvironment A: One AH representing flight dynamics for each

aircraft involved in the encounter T: One AH of the TCAS system for each aircraft E: One AH describes the airspace of the collision

encounter

AH - Aircraft

AH - TCAS

AH - Environment

TCAS Implementation - unmodified

Conforms to FAA specifications & algorithms

- Written in Visual Basic

- Uses FSUIPC 3rd party module by Pete Dowson

- Provides TCAS symbology

- Provides aural announcements

- Provides safe vertical speed ranges

EID-Enhanced TCAS Displays

D1: unmodified TCAS symbology

D2: circle around aircraft indicates protected volumes, red circle represents predicted collision area, time to loss of separation (LOS) is also indicated in seconds

D3: TIME to LOS is used as radar scale instead of separation distance, LOS time shown, and ground speed velocity indicators for each aircraft

Experimental Design

Participant background questionnaire TCAS, MSFS tutorial TCAS calculation proficiency exercise 3 sets of display condition trials (randomized)

Subjects press a button to show intent to maneuver 5 trials with colliding traffic, TCAS alerting on 8 trials with TCAS on/off, colliding/non-colliding traffic

evenly permutated scenarios (randomized order) NASA-TLX, self-confidence, and trust level surveys

Overall display preference questionnaire

Data Analysis

Within subjects design 1 factor (display condition: D1, D2, D3) repeated

measures analysis of variance (ANOVA) Dependent variables

Colliding traffic and TCAS alerting on Reaction time after alert until intent to maneuver Conformance to TCAS calculated maneuver

Colliding traffic and TCAS alerting off Reaction time to sense a conflict with intent to maneuver Conformance to theoretical TCAS calculated maneuver

Additional investigation of signal detection theory to determine effects across displays

Future Direction and Conclusions

Experimental results will indicate if EID-enhanced displays improve pilot reaction time and conformance to TCAS alerts

Results comparison between D2 and D3 will provide additional information on the effects of distance-scaled versus time-scaled displays on collision detection performance

Qualitative interpretation shall illustrate the influence of EID-enhanced displays on pilot trust in automated displays

Acknowledgements Centre for Research in Earth & Space Technology (CRESTech) Microsoft, software and hardware sponsor Dr. Catherine M. Burns Members of AIDL Jin Qian, Dr. Jeanette O'Hara-Hines

Department of Statistics & Actuarial Sciences Thesis readers

Dr. Carolyn G. MacGregor, SYDE Dr. Hamid Tizhoosh, SYDE

Q & A

Thank you!