Pedestrian Scenario Design and Performance Assessment in Driving Simulations

1
Pedestrian Scenario Design and Performance
Assessment in Driving Simulations

• Achal Oza, Qiong Wu, Ronald Mourant
• Virtual Environments Laboratory
• Northeastern University
• Presented by Achal Oza
• oza_at_coe.neu.edu
• Driving Simulation Conference 2005 North America
• November 30, 2005
• Orlando, FL

2
Overview

• Motivation
• Technology
• Triggering System
• Scenarios and Results
• Analysis

3
Motivation

• 4,749 pedestrians were killed in motor vehicle
  accidents in 2003
• Improve safety by developing scenarios that may
  be unsafe or impossible to test in a real-world
  environment
• Test subjects for situation awareness, hazard
  detection, and decision-making

4
Overview

• Motivation
• Technology
• Triggering System
• Scenarios and Results
• Analysis

5
Technology

• 91 Dodge Caravan
• Force feedback steering wheel and pedals
• Alienware computerPentium 4 3.4 GHzGeForce 6800
  Ultra
• Parabolic screen
• 1024 x 768 resolution30 horizontal field of view

6
Overview

• Motivation
• Technology
• Triggering System
• Scenarios and Results
• Analysis

7
Pedestrian Model

• MD2 file format (originally developed for Quake
  2)
• Two types of animations
• Walking
• Idling

8
Triggering System

• An objects action is triggered by another object
  performing a specified action
• Pedestrians use two type of triggers
• Traffic lights
• User vehicle

9
Triggering SystemTraffic Light Triggers

• The pedestrian is bound to a traffic light
• Walking is triggered by a green light
• Continues walking even after light change back to
  red

10
Triggering SystemUser Vehicle

• Pedestrian is bound to the user vehicle
• Walking is triggered when the vehicle enters a
  specified radius

11
Overview

• Motivation
• Technology
• Triggering System
• Scenarios and Results
• Crosswalk Scenario
• Intersection Dash Scenario
• Vehicle Turning Scenario
• Bus Stop Scenario
• Analysis

12
Crosswalk Scenario

• Crosswalk with a yield-to-pedestrians sign
• Crossing starts when driver is 225 feet away
• High pedestrian visibility

13
Results for Crosswalk Scenario

• 8 out of 10 subjects avoided a collision
• Stopped near crosswalk
• Good Time-To-Collision estimate
• The remaining two slowed down after collision

14
Overview

• Motivation
• Technology
• Triggering System
• Scenarios and Results
• Crosswalk Scenario
• Intersection Dash Scenario
• Vehicle Turning Scenario
• Bus Stop Scenario
• Analysis

15
Intersection Dash Scenario

• Driver waits at an intersection for a red light
  to turn green
• Pedestrian illegally crosses after the driver
  enters the intersection

16
Results for Intersection Dash Scenario

• 6 out of 10 subjects avoided a collision
• The remaining four ran through the red light

17
Overview

• Motivation
• Technology
• Triggering System
• Scenarios and Results
• Crosswalk Scenario
• Intersection Dash Scenario
• Vehicle Turning Scenario
• Bus Stop Scenario
• Analysis

18
Vehicle Turning Scenario

• Pedestrian crosses in front of the user vehicle
  while making a left turn at an intersection
• High pedestrian visibility

19
Results for Vehicle Turning Scenario

• 6 out of 10 subjects stopped before reaching the
  pedestrian
• Remaining four slowed down, without completely
  stopping, allowing the pedestrian to safely cross
  

20
Overview

• Motivation
• Technology
• Triggering System
• Scenarios and Results
• Crosswalk Scenario
• Intersection Dash Scenario
• Vehicle Turning Scenario
• Bus Stop Scenario
• Analysis

21
Bus Stop Scenario

• Initially obstructed from view by a bus
• Jaywalker enters street when driver is within 130
  feet
• Minimal pedestrian visibility

22
Results for Bus Stop Scenario

• 10 out of 10 subjects collided with the
  pedestrian
• 4 out of 10 stopped after the collision

23
Overview

• Motivation
• Technology
• Triggering System
• Scenarios and Results
• Analysis

24
Scenario Analysis

• Scenarios ranked from safe to dangerous
• Crosswalk
• Vehicle Turning
• Intersection Dash
• Bus Stop
• Safety increases when visibility increases

25
Future Study

• Focus on varying pedestrian visibility
• Larger sample sizes and more test groups
• Experienced and inexperienced drivers
• Add intelligence to the pedestrians
• Hesitate when crossing in a dangerous area
• Collect real-world data for validation (where
  possible)
• Use a variety of pedestrian models
• Children, the elderly, etc.

26
References

• 1. NHTSA, (2004). Traffic Safety Facts
  2003-National Highway Traffic Safety
  Administration
• 2. Kenneth R. Laughery, Theodore E. Anderson
  Edwin A. Kidd (1967). A computer simulation
  model of driver-vehicle performance at
  intersection. Proceeding of the 1967 22nd
  national conference. ACM Press. New York, NY,
  USA.
• 3. Dirk Helbing. (1992) Model for Pedestrian
  Behavior. Pages 93-98 in Natural Structures.
  Principles, Strategies, and Model in Architecture
  and Nature, Part II.
• 4. Staplin, L., K. Lococo, and S. Byington.
  Older Driver Highway Design Handbook. The
  Scientex Corporation Transportation Safety
  Division, Pennsylvania 19443, FHWA-RD-97-135,
  January 1998.
• 5. Gale, A.G., et al. Time-To-Collision As A Cue
  For Decision-Making in Braking. Vision in
  Vehicles. Vol. 3, 1991, pp. 19-26.
• 6. Hoffman, Errol R., and Rudolf G. Mortimer.
  Drivers Estimates Of Time To Collision. Accid
  Anal Prev, Vol. 26, No. 4, 1994, pp. 511-520.

27
Questions?

• Achal Oza
• oza_at_coe.neu.edu
• Virtual Environments Laboratory
• Northeastern University