Automated vehicle
Target Crash Population for Estimating the Safety Benefits of Automated Vehicle Functions Mikio Yanagisawa, Volpe National Transportation Systems Center July 22, 2015
Volpe The National Transportation Systems Center
2015 Automated Vehicles Symposium
Volpe The National Transportation Systems Center Advancing transportation innovation for the public good
U.S. Department of Transportation John A. Volpe National Transportation Systems Center
Research Purpose
Goal
Estimate potential safety benefits that could be gained from automated vehicle concept functions at NHTSA automation levels 2-4
Objectives
1. Map known automated vehicle functions and operations to crash information 2. Query national crash databases to estimate the target crash population that could benefit from automated vehicles
Focus
Light vehicles (Gross vehicle weight rating ≤ 10,000 pounds)
Research Approach 1. Describe automated vehicle functions 2. Identify/map target crash characteristics and determine: •
•
Target crashes that could be addressed by automated vehicle functions (L2-L4) in general Incremental target crashes that could not be addressed by crash-imminent avoidance systems (L0-L1)
3. Query and analyze crash data 4. Publish final report in 2015
Impact of Research Identification of the crash population that could benefit from automated vehicle functions First step in estimating potential safety benefits of L2-L4 automated vehicle functions Needs for future crash data collection
1. Describe Automated Vehicle Functions
NHTSA Automation Levels 0-1 L0 – No Automation
Driver is in complete and sole control of the primary vehicle controls at all times, and is solely responsible for monitoring the roadway and for safe operation of all vehicle controls.
Automation of one or more specific control functions operating independently. L1 – Function- Driver has overall control and is solely responsible for safe Specific operation, but can choose to cede limited authority over a Automation primary control. Vehicle can automatically assume limited authority over a primary control or provide added control to aid the driver
NHTSA Automation Levels 2-4 L2 - Combined Function Automation
Shared authority, driver cedes primary control but is still responsible for monitoring & safe operation. Driver is expected to be available at all times. Automation of at least two integrated primary control functions .
L3 - Limited Self- Driver can cede full control & monitoring authority under certain traffic & environmental conditions. Driver Driving Automation is expected to be available for occasional control. L4 - Full SelfDriving Automation
Driver provides destination or navigation input but is not expected to be available for control. Responsibility for safe operation rests solely on the automation system.
2. Identify/Map Target Crash Characteristics
Automated Vehicles and Crash Avoidance 350
Crash Unavoidable
300 250
Increasing Intensity of Action
Crash Threat Imminent
200 150 Normal Driving
100 50
Advisory Systems
0 0
Initial Threat
Driver Warning Systems Time
Automatic Control Intervention
173
Crash
Crash avoidance functions deal with target pre-crash scenarios in yellow/red zones Automated vehicle functions perform in green zone and may prevent entry in yellow/red zones → Potential reduction in exposure to driving conflicts
Layers of Crash Data Breakdown
Map automated functions to crash data Pre-Crash Modes Operational Conditions Causal Factors
Identify relevant variables and codes in General Estimates System (GES) and Fatality Analysis Reporting System (FARS) 5 distinct layers of crash data Pre-Crash Scenarios
Location • Highway • Non-Highway • Intersection Related • Ramp Related • Work Zone
• • • •
Run-Off-Road Crossing Paths Rear-End Opposite Direction • Lane Change • Use of Pre-Crash Typology
Driving Conditions • Surface • Lighting • Weather
Travel Speed • Low Speed • High Speed
Driver Condition • Driver Error • Recognition • Decision • Erratic • Physiological Impairment
Advantages of Crash Data Layers
Correlate automated level operational conditions to crash characteristics Automated functions only available on highways should correlate to crashes occurring on highways Technology may be limited to environmental or driver conditions
Identify overlapping target crash populations across automated level functions Functions can overlap; levels can overlap Can quantify target population by functions or aggregate to level of automation
Account for L0-L1 functions to avoid double counting
Pre-Crash Scenarios Backing 2%
Pedestrian Pedalcyclist 1% 1%
Opposite Direction 3% Animal 4% Lane Change 9%
Off Roadway 21% *Charts are used an example, this project will update these statistics with the 2013 crash databases
Other 4% Rear-End 29%
Crossing Paths 26%
W.G. Najm, B. Sen, J.D. Smith, and B.N. Campbell, Analysis of Light Vehicle Crashes and Pre-Crash Scenarios Based on the 2000 General Estimates System. DOT-VNTSC-NHTSA 02 04, DOT HS 809 573, February 2003
Rollover Sideswipe Vehicle Failure
Road Departure
Rear-End
Pedestrian
Other Parking
Opposite Direction
Object
Left Turn Across Path/ Opposite Direction (LTAP/OD) No Driver Non-Collision
Lane Change
Hit and Run
Evasive
Cyclist
Crossing Paths
Control Loss
Backing
Animal
Crash Type
Pre-Crash Scenario Animal/maneuver Animal/no maneuver Backing into vehicle Control loss/vehicle action Control loss/no vehicle action Turn right @ signal Straight crossing paths @ non signal Turn @ non signal Running red light Running stop sign Cyclist/maneuver Cyclist/no maneuver Evasive maneuver/maneuver Evasive maneuver/no maneuver Hit and run Turning/same direction Changing lanes/same direction Drifting/same direction LTAP/OD @ signal LTAP/OD @ non signal No driver present Non-collision - No Impact Object/maneuver Object/no maneuver Opposite direction/maneuver Opposite direction/no maneuver Other Parking/same direction Pedestrian/maneuver Pedestrian/no maneuver Rear-end/striking maneuver Rear-end/lead vehicle accelerating Rear-end/lead vehicle moving @ constant speed Rear-end/lead vehicle decelerating Rear-end/lead vehicle stopped Road edge departure/maneuver Road edge departure/no maneuver Road edge departure/backing Rollover Other - Sideswipe Vehicle failure
Distribution of Crash Causes (in % crashes) Atmospheric Visibility, 0.1
Vehicle Road Surface, 8 Factor, 2.5 Ill, 4.5 Driver Physiological condition
Asleep, 3.5 Recognition Error, 43.6
Drunk, 6 Erratic Action, 8.5
Automation Intervention Opportunity?
*Charts are used an example, this project will update these statistics with the 2013 crash databases
Decision Error, 23.3
Driver error
W.G. Najm, M. Mironer, J. Koziol, Jr., J.-S. Wang, and RR. Knipling, Synthesis Report: Examination of Target Vehicular Crashes and Potential ITS Countermeasures. DOT HS 808 263, June 1995
3. Query and Analyze Crash Data
Data Query and Analysis
Query 2013 GES and FARS databases Tabulate results in terms of (by level, by function): Target crash annual frequency, Fatal crash annual frequency, and Economic cost
Estimate incremental target crash population for automated vehicle functions Crash population not addressed by crash avoidance applications (L0-L1) Use available estimates of L0-L1 application effectiveness in crash avoidance
Volpe The National Transportation Systems Center
2015 Automated Vehicles Symposium
Volpe The National Transportation Systems Center Advancing transportation innovation for the public good
U.S. Department of Transportation John A. Volpe National Transportation Systems Center
Research Purpose
Goal
Estimate potential safety benefits that could be gained from automated vehicle concept functions at NHTSA automation levels 2-4
Objectives
1. Map known automated vehicle functions and operations to crash information 2. Query national crash databases to estimate the target crash population that could benefit from automated vehicles
Focus
Light vehicles (Gross vehicle weight rating ≤ 10,000 pounds)
Research Approach 1. Describe automated vehicle functions 2. Identify/map target crash characteristics and determine: •
•
Target crashes that could be addressed by automated vehicle functions (L2-L4) in general Incremental target crashes that could not be addressed by crash-imminent avoidance systems (L0-L1)
3. Query and analyze crash data 4. Publish final report in 2015
Impact of Research Identification of the crash population that could benefit from automated vehicle functions First step in estimating potential safety benefits of L2-L4 automated vehicle functions Needs for future crash data collection
1. Describe Automated Vehicle Functions
NHTSA Automation Levels 0-1 L0 – No Automation
Driver is in complete and sole control of the primary vehicle controls at all times, and is solely responsible for monitoring the roadway and for safe operation of all vehicle controls.
Automation of one or more specific control functions operating independently. L1 – Function- Driver has overall control and is solely responsible for safe Specific operation, but can choose to cede limited authority over a Automation primary control. Vehicle can automatically assume limited authority over a primary control or provide added control to aid the driver
NHTSA Automation Levels 2-4 L2 - Combined Function Automation
Shared authority, driver cedes primary control but is still responsible for monitoring & safe operation. Driver is expected to be available at all times. Automation of at least two integrated primary control functions .
L3 - Limited Self- Driver can cede full control & monitoring authority under certain traffic & environmental conditions. Driver Driving Automation is expected to be available for occasional control. L4 - Full SelfDriving Automation
Driver provides destination or navigation input but is not expected to be available for control. Responsibility for safe operation rests solely on the automation system.
2. Identify/Map Target Crash Characteristics
Automated Vehicles and Crash Avoidance 350
Crash Unavoidable
300 250
Increasing Intensity of Action
Crash Threat Imminent
200 150 Normal Driving
100 50
Advisory Systems
0 0
Initial Threat
Driver Warning Systems Time
Automatic Control Intervention
173
Crash
Crash avoidance functions deal with target pre-crash scenarios in yellow/red zones Automated vehicle functions perform in green zone and may prevent entry in yellow/red zones → Potential reduction in exposure to driving conflicts
Layers of Crash Data Breakdown
Map automated functions to crash data Pre-Crash Modes Operational Conditions Causal Factors
Identify relevant variables and codes in General Estimates System (GES) and Fatality Analysis Reporting System (FARS) 5 distinct layers of crash data Pre-Crash Scenarios
Location • Highway • Non-Highway • Intersection Related • Ramp Related • Work Zone
• • • •
Run-Off-Road Crossing Paths Rear-End Opposite Direction • Lane Change • Use of Pre-Crash Typology
Driving Conditions • Surface • Lighting • Weather
Travel Speed • Low Speed • High Speed
Driver Condition • Driver Error • Recognition • Decision • Erratic • Physiological Impairment
Advantages of Crash Data Layers
Correlate automated level operational conditions to crash characteristics Automated functions only available on highways should correlate to crashes occurring on highways Technology may be limited to environmental or driver conditions
Identify overlapping target crash populations across automated level functions Functions can overlap; levels can overlap Can quantify target population by functions or aggregate to level of automation
Account for L0-L1 functions to avoid double counting
Pre-Crash Scenarios Backing 2%
Pedestrian Pedalcyclist 1% 1%
Opposite Direction 3% Animal 4% Lane Change 9%
Off Roadway 21% *Charts are used an example, this project will update these statistics with the 2013 crash databases
Other 4% Rear-End 29%
Crossing Paths 26%
W.G. Najm, B. Sen, J.D. Smith, and B.N. Campbell, Analysis of Light Vehicle Crashes and Pre-Crash Scenarios Based on the 2000 General Estimates System. DOT-VNTSC-NHTSA 02 04, DOT HS 809 573, February 2003
Rollover Sideswipe Vehicle Failure
Road Departure
Rear-End
Pedestrian
Other Parking
Opposite Direction
Object
Left Turn Across Path/ Opposite Direction (LTAP/OD) No Driver Non-Collision
Lane Change
Hit and Run
Evasive
Cyclist
Crossing Paths
Control Loss
Backing
Animal
Crash Type
Pre-Crash Scenario Animal/maneuver Animal/no maneuver Backing into vehicle Control loss/vehicle action Control loss/no vehicle action Turn right @ signal Straight crossing paths @ non signal Turn @ non signal Running red light Running stop sign Cyclist/maneuver Cyclist/no maneuver Evasive maneuver/maneuver Evasive maneuver/no maneuver Hit and run Turning/same direction Changing lanes/same direction Drifting/same direction LTAP/OD @ signal LTAP/OD @ non signal No driver present Non-collision - No Impact Object/maneuver Object/no maneuver Opposite direction/maneuver Opposite direction/no maneuver Other Parking/same direction Pedestrian/maneuver Pedestrian/no maneuver Rear-end/striking maneuver Rear-end/lead vehicle accelerating Rear-end/lead vehicle moving @ constant speed Rear-end/lead vehicle decelerating Rear-end/lead vehicle stopped Road edge departure/maneuver Road edge departure/no maneuver Road edge departure/backing Rollover Other - Sideswipe Vehicle failure
Distribution of Crash Causes (in % crashes) Atmospheric Visibility, 0.1
Vehicle Road Surface, 8 Factor, 2.5 Ill, 4.5 Driver Physiological condition
Asleep, 3.5 Recognition Error, 43.6
Drunk, 6 Erratic Action, 8.5
Automation Intervention Opportunity?
*Charts are used an example, this project will update these statistics with the 2013 crash databases
Decision Error, 23.3
Driver error
W.G. Najm, M. Mironer, J. Koziol, Jr., J.-S. Wang, and RR. Knipling, Synthesis Report: Examination of Target Vehicular Crashes and Potential ITS Countermeasures. DOT HS 808 263, June 1995
3. Query and Analyze Crash Data
Data Query and Analysis
Query 2013 GES and FARS databases Tabulate results in terms of (by level, by function): Target crash annual frequency, Fatal crash annual frequency, and Economic cost
Estimate incremental target crash population for automated vehicle functions Crash population not addressed by crash avoidance applications (L0-L1) Use available estimates of L0-L1 application effectiveness in crash avoidance
