FHWA PedSafe: The San Francisco MTA/UC Berkeley Pedestrian Safety Program

Downloadable Version
PDF [1.07 MB]
PPT [7.72 MB]

Slide 1

FHWA PedSafe: The San Francisco MTA/UC Berkeley Pedestrian Safety Program

Frank Markowitz
SF Municipal Transportation Agency

David Ragland
UC Berkeley Traffic Safety Center

From TRB Annual Meeting, January 11, 2009

Slide 2

San Francisco Crosswalk

The San Francisco Setting

  • POPULATION: 800,000 in SF and 4.2 million in metro area (vs. Las Vegas 560,000/1.8 million and Miami 410,000/4.9 million – city/metro area)
  • HIGH PUBLIC TRANSIT USE: 9.5% Transit commuting in metro area (vs. about 4% for both Las Vegas and Miami)
  • PEDESTRIAN SAFETY INSTITUTIONALIZED by 2000 – fulltime pedestrian safety planning/engineering and outreach staff
  • OTHER – hills, multi-leg intersections, fog

Slide 3

Pedestrian Injuries

Line Chart: San Francisco Injury Collisions Involing Pedestrians (1998-2007) (Text version of chart below)

Figure 5
San Francisco Injury Collisions Involing Party Type Pedestrian (1998-2007)
Year 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Total 985 915 955 895 862 815 727 747 726 796

Return to top

Slide 4

Pedestrian Injury Patterns - Age

Bar Chart: Pedestrian/Driver Injury by Age (Text version of chart below)

Pedestrian Injury Patterns – Age
  0-4 5-9 10-14 15-19 20-24 25-44 45-64 65-74 75-84 75-84
Pedestrian 2.5 3.7 4.4 5.4 7.5 39.1 24.3 6.7 4.8 1.7
Driver     0 3.9 9.9 50.8 27.8 4.7 2.1 0.8

Return to top

 

Slide 5

Pedestrian Injury Patterns – Gender

Bar Chart: Pedestrian/Driver Injury by Gender (Text version of chart below)

Pedestrian Injury Patterns – Gender
  Female Male
Pedestrian
 45.5
 54.5
Driver 28.2 71.8

Return to top

 

Slide 6

Pedestrian Injury Patterns: Primary Violations

• Driver Failed to Yield to Pedestrian in Crosswalk35%
– At Signalized Crosswalk On Left Turn16%
• Pedestrian Failed to Yield, not Jaywalking13%
• Pedestrian Failed to Yield, Jaywalking
  (between signalized intersections)		9%
• Unsafe Speed7%
• Pedestrian Running in Crosswalk in front of vehicle7%
• Red Light Running3%

Return to top

 

Slide 7

PedSafe Zones

Map: San Francisco PedSafe (July 1996-June 2001)

Slide 8

Phase 1 – Planning: Recommended Countermeasures – 1

Pedestrian Safety Engineering Countermeasures

  • ADA Curb ramps and detectable warnings
  • Advance limit lines and red curb program
  • Distribution of retroreflective materials
  • Impactable and roadside "Yield to Pedestrian" signs
  • Median refuge island improvements
  • Pavement stencils ('Look both ways')
  • Pedestrian scramble phasing (exclusive pedestrian phasing)
  • Modified signal timing (increased pedestrian crossing time)
  • "Pedestrian head start" (leading pedestrian intervals)

Return to top

 

Slide 9

Phase 1 – Planning: Recommended Countermeasures – 2

Intelligent Transportation Systems (ITS) Technologies

  • Animated eyes/countdown pedestrian signals
  • Automated detection of pedestrians to extend crossing time
  • Flashing beacons (both automated detection and push button-actuated)
  • Pedestrian countdown signals (mostly replacing conventional pedestrian signals)
  • Portable radar speed trailer
  • Radar speed display sign
  • "Smart lighting" or other street lighting improvements

Return to top

Slide 10

Phase 1 – Planning: Recommended Countermeasures – 3

Education and Outreach Program:

  • In-person education in schools and in senior centers
  • Device-specific tips, including education at countermeasure sites
  • Grassroots media campaign

Return to top

Slide 11

Phase 2 – Implementation: Refining Countermeasures Plan

Refining Countermeasure List

  • Vendor interest in experimental measures
  • Cost considerations
  • Interagency negotiations

Assigning Countermeasures to Locations:

  • Primarily pedestrian or driver behavior problem?
  • Vehicle movement most frequently problem?
  • Type of violation?
  • Special opportunities?

Return to top

Slide 12

Results – Flashing Beacon with Automated Detection

Cost High
($62,000 installation cost)
Changes in MOES (pre vs. post-installation
Intersection 1
Ped. Delay 4.2 to 2.9 sec.
Ped. Look Before Crossing 87 to 95%
Diverted Crossing 19 to 5%


 

Vehicle Yield 81 to 94%
Vehicle/Ped Conflict 6 to 3%
Pedestrian Trapped 4 to 0%

Return to top

Slide 13

Results – Impactable Yield Sign

Cost Low ($1,800)
Intersection 4
Ped. Delay
Ped. Crossing Time		 5.1 to 4.6 sec. (NS)
11.0 to 11.0 sec.
Ped. Look Before Crossing
Ped. Look at Midpoint		 98.7 to 92.3%
70.3 to 56.0%
Diverted Crossings 14.6 to 16.0%
Vehicle Yield 52.6 to 67.6%
Vehicle/Ped. Conflict 7.2 to 6.8% (NS)
Pedestrian Trapped 3.7 to 4.3% (NS)

Return to top

Slide 14

Results- Ped Head Start (LPI)

Cost Low ($2,600)
Intersection 4
Ped. Delay 8.3 to 9.2 sec.
Ped. Start Crossing
% Red Hand Last 4 Sec
% Walk
% Flashing Red Hand
% Red Hand		  
10.2 to 7.5
66.0 to 71.5
16.9 to 18.2
6.2 to 5.9
% Veh turned Left in front of Ped.
% Veh turned right in front of Ped.FM1		 6.2 to 5.4 (NS)
6.6 to 5.7 (NS)
% Veh turned in
Front of Ped (3 intersections)		 6.2 to 4.0

FM1I had this combined vehicles turning in front of peds 6.2% pre vs. 4.0% post, which was significant. There was only 1 intersection I think in which RTs were even affected by the ped head start. I think it's a mistake to include turns that were not directly affected by head start.

Return to top

Slide 15

Results- Portable Speed Trailer

Cost Medium ($40,200)
Intersection 4
Ped. Delay 6.4 to 6.9 sec.
Ped. Crossing 14.0 to 13.9 sec. (NS)
Vehicle Yield 68.4 to 83.3%
Vehicle/Ped. Conflict
Veh. Speed ReductionFM2 		5.7 to 10.0 (NS)
Reduced 1 to 6 MPH

FM2I think Speed impacts are most important finding. That's what the device most directly targeted.

Return to top

Slide 16

Results – Automated Video Detection of Pedestrians to Extend Crossing Time

Cost High ($17,000 + Previously purchased equipment)
IntersectionFM4 1
% Peds. Finishing Crossing on Red 14 to 12 (NS)
% Late Crossing Peds Detected 100
% Extensions > 1 sec. 14
Extensions In Error
(due to vehicle encroachment)FM3		 5 in 30 minutes

FM3I think Speed impacts are most important finding. That's what the device most directly targeted.

FM4I don't think the MOEs for ped behavior are relevant because device was virtually invisible and not meant to influence ped behavior.

Return to top

Slide 17

Countdown Signals Program

Countdown Pedestrian Signal
  • Pilot – Countdown signals initially installed at 14 test locations
  • Full-scale Replacement – City eventually replaced most pedestrian signals with Countdown Signals (over 800 intersections)
  • More energy-efficient: electricity savings paying for LED device installation (over several years)

Return to top

Slide 18

Countdown Signals Impacts Summary:

  • Injuries – Citywide installation – about 22% drop in injuries at countdown locations vs. 2% rise at intersections without pedestrian signals but with traffic signals.
  • Pedestrians finishing crossing on red – significant reduction
  • Pedestrians starting to cross at the beginning of the clearance interval – no significant reduction
  • Red light running – no increase
  • Additional information on how much time left to cross the street liked by pedestrians.

Return to top

Slide 19

Countdown Behavioral Impacts

Countdown Behavior Impacts (Text version immediately below)

  Pre-Installation Post-Installation
Finish Crossing on Red 14% 9%
Run or Abort Crossing 13% 8%
Rating of Red Signal as "Very Helpful" 34% 76%

Return to top

Slide 20

Conclusions – Overview of most successful CMs

* Reported by SFMTA

Flashing Beacons

Impactable Yield Signs

Pedestrian Head Start

Portable Speed Trailer

Yield to Pedestrian Sign

Automatic Video Detection with Signal Extension*

Return to top

Slide 21

Overview of Data Collection and Analysis

  • Thirteen Countermeasures
  • 29 Intersections
  • >300 Hours of video
  • Intercept Surveys
  • Protocols Developed
  • Tools Used
  • Statistical Analysis
  • Inter-rater Reliability

FLASHING BEACON PROTOCOL – BASELINE

At the beginning of the clip, fill in the following variable names in the columns below the video screen (the number in front of each variable name correspond to the numbered instruction below):

Each independent pedestrian is counted separately (i.e., there are no groups). People being pushed in wheelchairs, strollers, etc, or small children accompanied by an adult are NOT independent. If the pedestrian starts across the street and then turns back, they are not to be counted. When the person being observed arrives at the intersection intending to cross, press the “New Initial Event” 1 button. A new row will begin and a time stamp will appear in the “Initial” column.

Ped Sweep

1. Ped Arrival 2. Ped Start 3. Gender 4. Age 5. % in Xwalk 6. Angel I/O 7. Ped Look
  1. When the pedestrian arrives at the corner of the crosswalk, press “Event” 1 button
  2. When the pedestrian starts to cross the street, press the “Event” 2 button. (This is when the pedestrian steps into the traffic lane. Stepping off curb into a parking lane does not count as “start to cross” unless the person does not pause in the parking lane.)
  3. Record the gender of the pedestrian. Enter a “9” if gender cannot be determined.
  4. Age (“Y” for youth – less than 21, “A” for adult – 21 to 60, or “E” for elderly – greater than 60). Enter a “9” if age cannot be determined.
  5. Record the percentage of time the pedestrian walks in the cross walk while crossing the street. (Use ONLY 25%, 50%, 7%, or 100% increments.)
  6. If the pedestrian is not in the cross walk 100% of the time, record if that person angled Into (I) the crosswalk as they began to cross or angled Out (O) of the crosswalk as they finished crossing.  If that person did both, record IO.
  7. Record if pedestrian looked for traffic before starting to cross (“1 if yes, “0” if no)

Return to top

Slide 22

METHODS

  • Use of Video Analysis
  • Signal Timing Linkage
  • Defining Operational Concepts (Interaction/Yielding/Conflicts)
  • Automated Analysis

Return to top

Slide 23

Use of Surrogate Measures & Video Observation

Video Observation in crosswalk
  • Limited time frame
  • Lack of collision data
  • Large amount of CMs vs. limited amount of intersections
  • Workload
  • The Playback Tool

 

 

 

 

 

Return to top

Slide 24

Lessons Learned – Video Analysis

Advantages

  • Repeated observations
  • Precise time stamping of events
  • Flexibility

Limitations

  • Labor-intensive
  • Storage requirement
  • Hardware requirement
  • Cost

Return to top

Slide 25

Signal Timing Linkage

Takes timestamps (from Play-Back Tool) and links actions to signal cycle

Signal Timing Linkage Chart (Text version immediately following chart)

The inputs of the pedestrian cycle are:

52.1
60.1
68.1
112.1

Pedestrain starts to corss the intersections

55.3
057.1
62.5

Pedestrian starts to walk in the Red Hand phase – Between 69-129 seconds

Knowing that the cycle is 60 seconds long and that the Walk phase began at 112.1 seconds, the computer knows the next Walk came at 172.1

Return to top

Slide 26

Signal Timing Linkage

Output:

Statistic Percentages for Each Phase
Howard and 8 Baseline Start Crossing     End Crossing
Red Hand-Last 4 Seconds 37 11.5%   Red Hand-Last 4 Seconds 0 0.0%
Walk   184 57.3%   Walk 4 1.2%
FRH   69 21.5%   FRH 176 54.8%
Red Hand-First 4 Seconds 1 0.3%   Red Hand-First 3.5 Seconds 41 12.8%
Red Hand-Up to the Last 4 Seconds 30 9.3%   Red Hand-3.5 to 4.5 Seconds 13 4.0%
Total   321 100.0%   Red Hand-Up to the Last 4 Seconds 69 21.5%
          Red Hand-Last 4 Seconds (Next Cycle) 1 0.3%
          Walk (Next Cycle) 17 5.3%
          Total 321 100.0%
 
Howard and 8 Baseline Start Crossing     End Crossing
Red Hand-Last 4 Seconds 37 11.3%   Red Hand-Last 4 Seconds 0 0.0%
Walk   182 55.5%   Walk 6 1.8%
FRH   73 22.3%   FRH 17 54.0%
Red Hand-First 4 Seconds 2 0.6%   Red Hand-First 3.5 Seconds 48 14.6%
Red Hand-Up to the Last 4 Seconds 34 10.4%   Red Hand-3.5 to 4.5 Seconds 9 2.7%
Total   328 100.0%   Red Hand-Up to the Last 4 Seconds 67 20.4%
          Red Hand-Last 4 Seconds (Next Cycle) 5 1.5%
          Walk (Next Cycle) 16 4.9%
          Total 328 100.0%

Signal Timing Linkage Statistic Pecentages for Each Phase Graph: Walk = 12, FRH = 13, Red Hand = 35, Cycle = 60
Graphs

Return to top

Slide 27

Video Observation Categories

Video Observation Categories: All Predestrians Crossing, Vehicle Interaction, Pedestrian Trapped, Vehicle Conflict, Vehicle Blockage, Vehicle Yield

Return to top

Slide 28

Vehicle/Pedestrian Interaction

Car a crosswalk and Pedestrian Crossing   Car at Crosswalk and Pedestrian Crossing

Return to top

Slide 29

Vehicle Yielding

Vehicle Yielding in Crosswalk  Vehicle Yielding in Crosswalk

Vehicle Yielding in Crosswalk

Return to top

Slide 30

Vehicle/Pedestrian Conflict

Vehicle/Pedestrian Conflict at Crosswalk  Vehicle/Pedestrian Conflict at Crosswalk

Vehicle/Pedestrian Conflict at Crosswalk  Vehicle/Pedestrian Conflict at Crosswalk

Return to top

Slide 31

Berkeley Traffic Scene Analysis System

  • Advanced algorithm combining background subtraction and feature tracking
  • Pedestrian/vehicle detection and tracking
Berkeley Traffic Scene Analysis System Diagram 1
Berkeley Traffic Scene Analysis System Diagram 1 		Berkeley Traffic Scene Analysis System Diagram 1
Berkeley Traffic Scene Analysis System Diagram 1

Return to top

Slide 32

Berkeley Traffic Scene Analysis System (Continued)

  • Vehicle/non-vehicle classification

Berkeley Traffic Scene Analysis System - Vehicle/non-vehicle classification  Berkeley Traffic Scene Analysis System - User-assisted system

  • User-assisted system was also developed to obtain perfect result

Return to top

Slide 33

Berkeley Traffic Scene Analysis System (Continued)

Berkeley Traffic Scene Analysis  Berkeley Traffic Scene Analysis

Berkeley Traffic Scene Analysis  Berkeley Traffic Scene Analysis

Return to top

Slide 34

Next Steps: Additional Research

  • ITE Committee Preparing Information Report on Automated Detection Technologies
  • Surveying Effectiveness of Different Devices, but also:
    • Costs
    • Maintenance
    • Liability Issues
  • Timely with MUCTCD Changes
    • Especially extended crossing time
  • More Concret Definitions
  • Automated Pedestrian Counting

Return to top

 

 

Program Contact

Tamara Redmon

202-366-4077

Dick Schaffer

202-366-2176

What’s New

The FHWA Safety Office is continually developing new materials to assist states, localities and citizens in improving pedestrian and bicycle safety. The materials listed on this page were completed recently.

New Spotlight on Pedestrian Safety

New Promoting the Implementation of Proven Pedestrian Countermeasures

Pedestrian Forum - Fall 2011

State Best Practice Policy for Medians

State Best Practice Policy for Shoulders and Walkways

Pedestrian Countermeasure Policy Best Practice Report

The State of Florida is developing a statewide Pedestrian Safety Action Plan. They have set up a project website that includes information about the project, workshop presentations and resources relating to pedestrian safety.

Evaluating Pedestrian Safety Countermeasures

Safety Benefits of Raised Medians and Pedestrian Refuge Areas: Brochure, Booklet

Safety Benefits of Walkways, Sidewalks, and Paved Shoulders: Brochure, Booklet

Pedestrian Safety Strategic Plan