SAfety VEhicles using adaptive Interface Technology
This presentation is the property of its rightful owner.
Sponsored Links
1 / 22

Overview Gerald Witt & Harry Zhang August 12, 2003 PowerPoint PPT Presentation


  • 83 Views
  • Uploaded on
  • Presentation posted in: General

SAfety VEhicles using adaptive Interface Technology Phase 1 Research Program Quarterly Program Review. Overview Gerald Witt & Harry Zhang August 12, 2003. SAVE - IT Phase 1 Program Overview. Program Team Mission and Objectives Program Plan Summary Technical Approach

Download Presentation

Overview Gerald Witt & Harry Zhang August 12, 2003

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Presentation Transcript


Overview gerald witt harry zhang august 12 2003

SAfety VEhicles using adaptive Interface Technology Phase 1 Research ProgramQuarterly Program Review

Overview

Gerald Witt & Harry Zhang

August 12, 2003


Save it phase 1 program overview

SAVE - IT Phase 1 Program Overview

  • Program Team

  • Mission and Objectives

  • Program Plan Summary

  • Technical Approach

    • Phase 1 Research model

    • Team coordination

    • Schedule

    • Human Factors research summary


Program team

Program Team

A comprehensive program team has been assembled bringing a unique blend of expertise and complimentary capabilities.

Program Manager

Gerry Witt

DDE Human Factors

Team Leader

Dr. Harry Zhang

DDE Technology

Team Leader

Greg Scharenbroch

DDE

Principal Investigator

Greg Scharenbroch

Seeing

Machines Inc.

Ford

Evaluation

Principal Investigator

Jeff Greenberg

U of Iowa

Principal Investigator

Dr. John Lee

UMTRI

Principal Investigator

Dr. David Eby

DDE

Principal Investigator

Dr. Harry Zhang

GM

Evaluation

Principal Investigator

Scott Geisler

  • Task Members

  • Tim Newman

  • Dr. Branislav Kisacanin

  • Nancy Edenborough

  • Michelle Wilkes

  • Development Focus

  • Technology

  • Development

  • System

  • Integration

Eye Tracking

Technology

  • Task Leader

  • Jeff Greenberg

  • Research Focus

  • Evaluation

  • Task Leader

  • Scott Geisler

  • Research Focus

  • Evaluation

  • Task Leaders

  • Dr. John Lee

  • Dr. Dan McGehee

  • Dr. Tim Brown

  • Research Focus

  • Distraction Mitigation

  • Cognitive Distraction

  • Telematics Demand

  • Guidelines and

  • Standards

  • Evaluation

  • Task Leaders

  • Dr. David Eby

  • Dr. Paul Green

  • Dr. Bary Kantowitz

  • Dr. Dave LeBlanc

  • Research Focus

  • Scenario ID

  • Driving Task Demand

  • Performance

  • Telematics Demand

  • Evaluations - on road

  • Task Leaders

  • Dr. Zhang

  • Dr. Smith

  • Research Focus

  • Visual Distraction

  • Intent

  • Safety Warning

  • Countermeasures

  • Data fusion

  • Benefits Analysis


Save it mission and objectives

SAVE - IT Mission and Objectives

Develop system

operational performance

requirements and guidelines

for adaptive interface

conventions

Conduct comprehensive

human factors research to derive

distraction and workload

measures for use adaptive

interfaces.

Identify scalable system

concepts and sensing

technologies for further

research to follow

the SAVE-IT program

Phase 1

E

C

T

J

I

B

V

E

O

Mission

To demonstrate a viable proof of

concept that is capable of reducing

distraction related crashes and

enhancing collision warning

effectiveness

S

Phase 1

Phase 1

Develop and apply

evaluation procedures

for assessment of

safety benefits

Phase 2

Provide the public

with documentation on

human factors research

findings for performance

and standardization

development

Phase 1

Advance the

deployment of adaptive

interface technology

countermeasures for

distraction related

crashes

Enhance collision

warning effectiveness by

optimizing alarm onset based

on driver’s workload

or distraction


Program plan summary

IOWA

UMTRI

Program Plan Summary

Phase II

Data Fusion, System Integration and Evaluation

Phase I

Research and Concept Development

Evaluation

14A Iowa

14B Ford

14C UMTRI

14D GM

14

Scenario

Identification

Data Fusion

11ADistraction

Mitigation

11BSafety Warning

Countermeasures

Subcontractors:

11

System

Integration

Vehicle

build

Demo.

13

Crash statistics analysis

1

Driving Task

Demand

Literature review

2A

Identify diagnostic measures

2B

Develop and validate algorithms

2C

Literature review

3A

Identify diagnostic measures

3B

Develop and validate algorithms

3C

Performance

Distraction

Mitigation

Literature review

Cognitive distraction

Visual distraction 4A

Identify countermeasures

Cognitive distraction

Visual distraction 4B

Validate countermeasures

Cognitive distraction

Visual distraction 4C

Cognitive

Distraction

Literature review

5A

Identify diagnostic measures

5B

Develop and validate algorithms

5C

Telematics

Demand

Literature review

6A

Identify demand levels

6B

Validate demand levels

6C

Iowa

Visual

Distraction

Literature review

7A

Identify diagnostic measures

7B

Develop and validate algorithms

7C

UMTRI

Literature review

8A

Identify diagnostic measures

8B

Develop and validate algorithms

8C

Intent

Program Summary

and Benefit Evaluation

15

Literature review

9A

Identify countermeasures

9B

Safety Warning

Countermeasures

Establish

Guidelines &

Standards

12

Technology / architecture

concept identification 10A

Technology / architecture

concept car 10B

Technology

Development

2004 - 2005

2003

DELPHI


Save it phase 1 research model

SAVE-IT Phase 1 Research Model

Technology/Architecture Concept

Development

Human Factors

Research

Concept

Demonstration

Adaptive Safety Warning

And Distraction Mitigation

System Architecture Concepts

Scenario

Identification

Diagnostic Research

Driving Performance

Real Time Distraction

Sensing Requirements

Cognitive Distraction

Visual Distraction

Phase 2

Recommendations

and development

plan

Distraction Assessment

Data Fusion Concepts

Telematics Demand

Driving Task Demand

Situational Threat

Assessment Concepts

Intent

CountermeasureTechnology Identification

and HMI Concepts

Safety Warning

Countermeasures

Distraction Mitigation


Preliminary save it model

Preliminary SAVE-IT Model

  • Provides real time assessment of driver distraction

  • Provides global situational threat assessment

  • Provides adaptive countermeasures.

  • Provides a watchful eye when your not

  • Increases safety guard band when required

Target

Steering

Assessment

Throttle

Pedestrian Detect

HMI data fusion Processor

Brakes

FW Long Range

FW Short range

Phone

Side Detect

Vehicle Control

Longitudinal

MMM

N

Lateral

R Long Range

Response

Req’d

Climate

Situational

R. Short Range

Threat

Y

3D Audio

Assessment

Adaptive HMI

Environment

Stimulate/Suppress

HUD

Warning sensitivity

Driver

State

Monitor

MMM

Information priority

Haptic FB

Phone

Displays

IP Controls

Flashers

Warnings

Steering

CHMSL

Throttle

  • Substantially reduces perceived

  • false alarm conditions and minimizes

  • driver disregard.

Bio Signs

Brakes

Eye Tracking/

Oculometrics


Team coordination

Team Coordination

  • Close team coordination is required to maintain consistency within research, experimental design and conclusions

  • A bi- weekly Human Factors team meeting is held via conference call.

    • Schedule review

    • Design Reviews

    • Issue discussions and resolutions

    • Commonization strategy discussion

      • Common dependant variables

      • Age groups, etc.

  • Common development process

    • Literature review

      • Report

    • Design/data collection

      • Team design reviews

      • IRB approval

      • Data collection

      • Findings and recommendations

    • Phase 2 planning

      • Algorithm development and validation plan

      • Preliminary Phase 2 research plan


Phase 1 schedule milestones

Phase 1 Schedule Milestones

  • Literature reviews complete, report submitted to NHTSA/Volpe9/10/2003

  • Final Reports and Recommendations to Delphi12/31/2003

  • Phase 2 planning documentation to Delphi12/31/2003

  • Final report and recommendations to NHTSA/Volpe3/4/2004

  • Phase 2 concept vehicle demonstration to NHTSA/Volpe3/4/2004

  • Phase 2 planning documentation to NHTSA/Volpe3/4/2004


Overview gerald witt harry zhang august 12 2003

Comparison of Approaches

Demand Approach

COMUNICAR

(Europe)

Non-Adaptive-

Interface Approach

GIDS

(Europe)

Arousal Approach

SAVE-IT:

Real-time

Adaptive interface

Common scenarios among tasks

Driving & non-driving demands

Driver state (distraction,

intent, physiological measures)

Safety warning systems

CAMP Workload

(U.S., Japan)

De Waard

Comprehensive Safety Management Systems


Philosophy of comprehensive safety management systems

Philosophy of Comprehensive Safety Management Systems

  • Driver impairment reduces overall attentional capacity.

  • Driver distraction increases the attention allocated to non-driving tasks and reduces the attention allocated to driving tasks.

  • Safe driving requires commensurate attention paid to driving tasks.

  • Required attention to driving tasks varies with driving task demand.

  • Objectives: To assess distraction, impairment, and driving task demand in order to ensure sufficient attention is paid to driving tasks.


Task 1 crash statistics analysis

Task 1: Crash Statistics Analysis

  • Objectives

    • Identify crash scenarios (e.g., rear-end crashes) that the SAVE-IT program should be designed to prevent.

  • Major Findings

    • 20-50% of crashes involve some form of driver distraction and inattention.

    • CDS appears to be best suited for the task. FARS, GES, and HSIS are not appropriate for this task.

    • Prior research indicated that single-vehicle-run-off-the-road and rear-end crashes are most common scenarios in which driver distraction is a causative factor.

    • Prior research indicated distracting events include interior (e.g., radio, cell phones, passengers) and exterior (e.g., scenery) objects and events.

  • Current Status(55% completed)

    • Literature review report (to Delphi)Completed

    • Crash data analysis planIn progress

    • Crash data analysisSept.-Nov. ’03

    • Expert panel meetingNov. ‘03


Task 2 driving task demand

Task 2: Driving Task Demand

  • Objectives

    • Determine the level of attentional demand imposed by the driving environment that represents the required level of attention allocated to the driving tasks.

  • Major Findings

    • Analysis of crash data is key because crash rates can be assumed to indicate environmental unpredictability and the amount of attention demanded by the environment.

    • HSIS is the best suited crash database because it contains information about environmental conditions (e.g., weather, traffic volume, road surface conditions) at the time of crashes.

    • In laboratories, driving demand can be approximated by visual demand (% of time needed to look at the road to drive safely) as measured with the visual occlusion method.

  • Current Status (2A 50% completed; 2B 60% completed)

    • Literature review report (to Delphi) (2A)Completed

    • HSIS database preparation, review of analysis plan (2A)Completed

    • Crash data analysis (2A)Aug.-Sept. ’03

    • Review of test plan, simulator preparation, pilot testing (2B)Completed

    • “Visual occlusion” simulator experiment (data collection) (2B)In progress

    • Data analysis and algorithm development (2B)Sept-Nov. ’03


Task 3 performance

Task 3: Performance

  • Objectives

    • Determine performance measures/variables that are diagnostic of driver distraction.

  • Major Findings

    • Literature review indicated that there exists very limited data (e.g., distribution data, eye glance data) comparing driving with and without various in-vehicle devices (e.g., radio, phone, navigation device).

    • NHTSA’s 100-car naturalistic driving study currently conducted at Virginia Tech should be very useful.

    • Normative data on drugs and driving can be useful.

  • Current Status (3A 15% completed; 3B 55% completed)

    • Identification of research needs/gaps (3A)Completed

    • Literature review report (to Delphi) (3A)In progress

    • Instrumented car preparation, pilot testing (3B)Completed

    • Review of test plan (3B)Completed

    • On-road experiment (data collection) (3B)Summer ’03

    • Data analysis and algorithm development (3B)Sept-Dec. ’03


Task 4 distraction mitigation

Task 4: Distraction Mitigation

  • Objectives

    • Develop appropriate countermeasures that can mitigate against excessive levels of distraction.

  • Major Findings

    • Research on computer etiquette, negotiated access, and automation challenges is useful.

    • Used four focus groups (24 participants at Iowa City & Seattle) to determine what activities drivers find distracting (driver/system initiated, technology/non-technology oriented) and what mitigation strategies they prefer. Potential technology for mitigating distraction is viewed positively by some and negatively by others.

    • Potential countermeasures such as warning, informing, advising, demand minimizing, prioritizing/filtering, locking, etc. can be summarized by a model-based taxonomy of mitigation strategies with degree of intervention and locus of control (driver vs. IVIS) as the dimensions.

  • Current Status (45% completed)

    • Literature review report (to Delphi) (4A)In progress

    • Focus group study (data collection)(4B)Completed

    • Focus group data analysis and draft report (4B)Completed

    • Revision of mitigation taxonomy based on focus group input (4B)Completed

    • Cognitive task analysis (4B)Aug.-Dec. ‘03

    • “Driver acceptance” simulator experiment (4B)Aug.-Dec. ‘03


Task 5 cognitive distraction

Task 5: Cognitive Distraction

  • Objectives

    • Determine which measures (performance, driver state, and vehicle state variables) are diagnostic of cognitive distraction.

    • Develop an algorithm relating diagnostic measures to performance (including RT).

  • Major Findings

    • Cognitive distraction may be manifested in terms of driver state (eye movements, scan patterns, ocular responses, psycho-physiological measures), driving performance, and vehicle system state.

    • Theories and models such as multiple resource theory, malleable resource theory, strategic task management (switching), and ACT-R can be useful.

    • Hidden Markov Models (representing stochastic sequences where states are not directly observed but are associated with a probability density function) and Support Vector Machines (determining optimal hyperplane separating two classes) will be very useful in predicting driver distraction.

  • Current Status (40% completed)

    • Literature review report (to Delphi) (5A)In progress

    • Experimental design (5B)In progress

    • Simulator experiment (5B)Aug.-Dec. ‘03


Task 6 telematics demand

Task 6: Telematics Demand

  • Objectives

    • Determine distraction potential and prioritization for commonly-used telematics functions.

  • Major Findings

    • Distraction potential may be measured in terms of task completion time, number of errors, number of glances, mean glance duration, reaction time, etc.

    • Current guidelines (e.g., Alliance’s Statement of Principles, SAE J2364) and IVIS Demand model for key task characteristics (visual, auditory, cognitive, manual) can be very useful.

  • Current Status (6A 60% completed, 6B 2% completed)

    • Summary of prior research (6A)Nearly completed

    • Literature review report (to Delphi) (6A)In progress

    • Preliminary test plan (6B)Completed

    • Simulator experiment (6B)Sept.-Dec. ‘03


Task 7 visual distraction

Task 7: Visual Distraction

  • Objectives

    • Identify eye glance measures that are diagnostic of visual distraction and that can be used in real-time, adaptive interface technology systems.

    • Determine performance (including RT) effects of visual distraction. For example, RT = f(glance duration, glance frequency, etc.).

  • Major Findings

    • Prior research indicated that visual distraction (off-road glances) degrades driving performance (e.g., SDLP, lane departures, RT) and increases the likelihood of crashes.

    • Prior research rarely used automatic eye tracking systems to measure visual behaviors in real time and focused on task-based (e.g., radio tuning) rather than time-based visual behaviors.

    • Real-time measurement of time-based visual behaviors is critical to SAVE-IT.

    • Current experiments measured visual behaviors in real time using Seeing Machines eye tracking system and indicated that visual distraction degraded performance (e.g., RT).

  • Current Status (7A completed; 7B 40% completed)

    • Literature review report (to Delphi) (7A)Completed

    • Experiment 1 data collection (7B)Completed

    • Review of Experiment 2 test plan, facility preparation (7B)Completed

    • Experiment 2 data collection (7B)In Progress

    • Data analysis, algorithm development (7B)Sept-Dec. ’03


Task 8 intent

Task 8: Intent

  • Objectives

    • Identify a list of Intents that are measurable and potentially useful for distraction mitigation and safety warning countermeasures.

    • Determine diagnostic measures for reliable detection of those intents.

  • Major Findings

    • Information about driver intent can improve system effectiveness and acceptance and reduce nuisance alerts. For example, FCW warnings may be delayed/suppressed if drivers intend to brake, and blind spot warnings may be issued when drivers intend to change lane.

    • Intent detection variables can be classified into affordance (e.g., exit ramp), motive (e.g., navigation info), kinematics (e.g., raw, heading), control (e.g., turn signal), and eye glances.

  • Current Status(50% completed)

    • Literature review report (to Delphi) (8A)Completed

    • Framework for intent determination (8B)Completed

    • Acquisition and preliminary examination of ACAS FOT pilot data (8B)Completed

    • Naturalistic Lane Change Data

      • Markov matrix analysis of eye movements (8B)Completed

      • Analysis of kinematic & control variables (8B)In progress


Task 9 safety warning countermeasures

Task 9: Safety Warning Countermeasures

  • Objectives

    • Improve the effectiveness and acceptability of safety warning systems by designing these systems to adaptively respond to intent, distraction, and demand information.

  • Major Findings

    • Although adaptive systems can fail because of poor user acceptance (e.g., not understanding the system, perceived system inconsistency or unpredictability, drivers feeling no longer in control) and poor design (e.g., oscillations), they have shown some promise and can be acceptable to drivers.

    • Excessive nuisance alerts pose major problems for FCW and other warning systems. Nuisance alerts may be reduced by adjusting warning criteria using driver state information (distraction and intent).

      • FCW warnings and lane drift warnings may be delayed/suppressed if drivers are attentive or intending to engage in particular maneuvers.

      • Blind spot warnings may be issued when drivers intend to change lane.

  • Current Status (45% completed)

    • Literature review report (to Delphi) (9A)Completed

    • Identification and definition of adaptive enhancement issues (9B)Completed

    • Selection of FCW algorithm and DVI (9B)Completed

    • Preliminary design of experiments (9B)Completed

    • 2 simulator experiments (9B)Sept.-Dec. ‘03


Common issues and solutions

Common Issues and Solutions

  • Regular team meetings convened to

    • Discuss common issues and solutions.

    • Commonize design and variables across experiments as much as possible.

  • Dependent Variables

    • Four common dependent variables used across all experiments

      • Reaction time for initial brake application

      • Reaction time for foot off accelerator

      • Steering entropy

      • Steering reaction time and direction

    • Many other variables (e.g., SDLP, lane departures, head and eye movements) will be used, although they may vary in different experiments.

  • Subject Ages

    • Subject ages divided into 3 age groups: 18-25; 35-55; and 65-75 years old.

    • All experiments must include the group of 35-55 years old.

    • Due to time and budgetary constraints, there is no requirement to use the younger (18-25 years old) or older subjects (65-75 years old). If multiple age groups are used, however, the additional group(s) may be 18-25 years old, or 65-75 years old, or both.

    • The choice of subject ages is made to balance the age effect and age representation (e.g., inclusion of subjects of 45 years of age or older). 35-55 years old are also the likely initial adopters of SAVE-IT technologies.


Common issues and solutions1

Common Issues and Solutions

  • All simulator experiments (at Delphi, Iowa, and UMTRI) use GlobalSim Simulator.

  • All visual glance behaviors measured with Seeing Machines eye tracking system.

  • If applicable, the driving scenario is the “lead vehicle following” scenario.

    • The same vehicle (white passenger car) used as the lead vehicle in all experiments.

    • For all experiments, a “rubber-band” control (Lee et al., 2002) is used to set time headway at 1.8 s at the moment of lead vehicle braking.

    • Except for Task 9, lead vehicle braking is non-imminent (e.g., -0.2 g).

    • Lead vehicle braking is unpredictable and infrequent (e.g., less than 1 per minute).

  • If practical, both rural roads and freeways are used.

    • Target speed = 45 mph for rural roads and 65 mph for freeways.

    • Some experiments use one type of roads only because the use of both types of roads may result in excessive long sessions or the use of too many subjects.

  • Per Volpe’s request, the Hidden Markov Model will be considered as a method in Tasks 3 (performance), 5 (cognitive distraction), and 8 (intent).

  • The minimum number of subjects per condition is 8. Many more subjects are used in many experiments.


  • Login