Intersection Decision Support for the Driver Minnesota Team

Intersection Decision Support for the Driver Minnesota Team: Mitigation of Crashes At Unsignalized Rural Intersections 1 st CICAS Coordination Meeting Sept. 27, 2004

National Motivation u 2. 7 million annual intersection related crashes u Represents 42. 7% of all 6. 32 million police reported crashes v 48. 1% occurred at signalized intersections v 52. 1% occurred at unsignalized intersections (stop sign, no controls, other sign). u 9, 760 of 38, 309 (25. 4%) of fatal crashes were intersection related …. . Traffic Safety Facts 2002

Intersection Decision Support (IDS) u Focus on driver error causal factors v Minnesota model: • Fatal intersection crashes • Provide the driver with information that will improve judgment of gap clearance and timing

Crossing Path Crash Causal Factors at Intersections (1998 GES) From Najm W G, Koopmann J A, and Smith D L (2001) Analysis of Crossing Path Crash Countermeasure Systems. Proc. 17 th Intl Conference on Enhanced Safety of Vehicles Left turn Right angle Right turn Traffic Cntrl Device Signal Stop Sign No Controls LTAP/OD: Left turn across path/ opposite direction LTAP/LD: Left turn across path/ lateral direction LTIP: Left turn into path RTIP: Right turn into path SCP: Straight crossing paths Virtually no crashes in shaded cells. Causal Factor Insuf. Gap Signal Viol. Insuf. Gap Sign Viol. Insuf. Gap Crossing Path Pre-Crash Scenarios LTAP/OD LTAP/LD LTIP RTIP SCP 193, 000 13, 000 31, 000 52, 000 15, 000 6, 000 178, 000 15, 000 113, 000 26, 000 25, 000 173, 000 12, 000 7, 000 3, 000 62, 000 92, 000 25, 000 10, 000 11, 000 35, 000

Minnesota Focus: Rural Unsignalized Intersections u u u Crashes in rural areas are more severe than in urban areas v While 70% of all crashes in Minnesota occur in urban areas, 70% of fatal crashes occur in rural areas. Along Minnesota’s Trunk Highway System, there are more rural through/stop intersections (3, 920) than all categories of urban intersections (3, 714) combined During a three-year period (1998 -2000), 62% of all intersection-related fatal crashes in Minnesota occurred at rural through/stop intersections

Addressing Rural Intersection Safety Issues: u The primary problem at rural intersections involves a driver on the minor road selecting an unsafe gap in the major road traffic stream. u Consider study of 1604 rural intersections (2 -lane roadways, Thru/STOP intersection control only, no medians) over 2+ year period.

Addressing Rural Intersection Safety Issues u Analyzed 768 right angle crashes on 409 different intersections. u Nearly 60% occur after vehicle on the minor roadway stops u Approximately 25% involved vehicle running through the STOP sign. … i. e. Problem is one of gap selection, NOT intersection recognition Source: Howard Preston CH 2 MHill

Minnesota Components u u u Rural unsignalized intersections: v High-speed corridors v Through stop intersections Traffic surveillance and wireless communications technologies (& on-site validation) Gap detection/estimation (& on-site validation) Human interface design and evaluation (in driving simulator) Goal - Results from above activities to lead to national Field Operational Test: v Application to MUTCD National Committee for DII approval

Focused on Recognized National Problem Guidelines for Implementation of AASHTO Strategic Highway Safety Plan u NCHRP Report 500: Vol. 5 Unsignalized Intersections v Identifies objectives and strategies for dealing with unsignalized intersections v Objective 17. 1. 4 Assist drivers in judging gap sizes at Unsignalized Intersections v High speed at grade intersections

Intersection Decision Support Program Tasks u u u Crash analysis; site selection Enabling research v Surveillance system deployment and evaluation v Driver Infrastructure Interface (DII) design and evaluation Benefit: Cost analysis Prediction of countermeasures effects System design

Identification of Experimental Site: Minnesota Crash Data Analysis 3, 784 Thru-STOP Isxns in MN Hwy System were evaluated Total >CR (% of total) 2 -Lane - 3, 388 | 104 (~ 3%) Expressway - 396 | 23 (~ 6%) Howard Preston, Richard Storm, Max Donath, Craig Shankwitz "Review of Minnesota's Rural Intersection Crashes: Methodology for Identifying Intersections for Intersection Decision Support (IDS)" (May 2004) Located at: http: //www. its. umn. edu/research/applications/ids/consortium/publications/ index. html

Thru-STOP Rural Expressways have Higher % of Right Angle Crashes u Most other crash types Source: Mn/DOT 2000 – 2002 Crash Data show decrease. u % right angle crashes significantly increases for rural expressway isxns over critical crash rate. u Rural expressway isxn over critical crash rate (Hwy 52 & CSAH 9) selected as test site. u Selected isxn (ADT>17, 500) was programmed into driver simulator for DII eval. u All vehicles passing thru isxn ROI will be tracked

Candidate Intersections: At-Fault Driver Age Source: Mn/DOT 2000 – 2002 Crash Data

Location of Selected Intersection MN Hwy 52 & CSAH 9

Sight distance restricted on the W approach at CSAH 9 Note differences in N and S vertical alignments

Approach u Measure gaps that drivers take under actual road conditions. Collect data regarding existing intersection entry behavior, and document the difference between a safe and unsafe gap (related to near misses and crashes). u Evaluate suite of sensors to ensure that they are able to measure those gaps accurately and able to distinguish between the unsafe and safe gaps at the level that the existing literature specifies. Use existing but limited literature (AASHTO standards) to quantify safe vs unsafe gap u Determine a set of interfaces with which to communicate the existence of an unsafe vs safe gap in the intersection at the moment the driver is making his or her decision. u Evaluate the selected set of interfaces in a driving simulator under "stressing", i. e. challenging conditions. (Use established safe vs unsafe gap definitions. )

Surveillance System - Overview u u u System designed to record the location and velocity of every vehicle at or approaching the intersection Surveillance system consists of an array of sensors v Radar, Lidar (LIght Detection And Ranging) v Vision – visible and infrared, image processing Sensor data transmitted to central processor v Sensor data filtered and fused v Intersection vehicle state matrix v Gaps in traffic calculated v Warnings generated for Driver Infrastructure Interface (DII) Communications (Wired / Wireless options) v Transmit data from sensors to IDS main processor (RSU) Performance issues: v Redundancy, reliability, range, power, cost, estimation vs. sensor coverage, etc.

Test Intersection Sensor Configuration: Major Leg – Hwy 52 v Approximately 2100 ft of lane coverage by radar in each direction – track 17. 2 sec. at 85 mph v Average sensor orientation angle is 4. 9º

Radar Performance Evaluation: Geometrics

Experimental Setup: Radar Now Picks up Vehicles at 440 ft.

Experimental Setup: Playing back experimental data

LIDAR Detectors u u u SICK LMS 221 sensors – operate at 5 Hz; for low speed minor leg Developed roadside vehicle detection (how many? ) and classification algorithm Calibration/testing to be completed in Oct. 2004

Test Intersection Sensor Configuration: Minor Leg – Goodhue CSAH 9 v v Radar and lidar sensors on CSAH 9 Radar to detect approaching traffic and lidar used for slow/stopped traffic Vehicle classification radar and lidar also used Queue length measured on collector

Cameras on masts u Multiple camera (& IR) / mast units provide independent means to verify: v Radar performance v Camera performance v Gap tracker performance v Other emerging technologies u May or may not be part of final system. Decision based on v Benefit: cost analysis v Poor weather performance of vision based systems

Human Factors Tasks u Analyze problem v Task analysis • “What are drivers doing wrong? ” • “Who is at most risk? ” v Driver model (Information Process) • “Why are they doing it wrong? ” • “What information could support correct behavior? ” v Previous solutions • “What has not worked before? ” u Simulate case site Propose interfaces and simulate candidate Evaluate candidate interfaces u u

Target Population u Older drivers (> 65 years) have a high crash risk at intersections Drivers > 75 years had greatest accident involvement ratio (Stamatiadis et al. , 1991) v Drivers > 65 years 3 to 7 times more likely to be in a fatal intersection crash (Preusser et al. , 1998) v Drivers > 65 years over-represented in crashes at many rural intersections in Minnesota (Preston & Storm, 2003) v

Intersection Simulation Task

Variable Message

Several Prototypes Hazard Beacon • Flashing sign activates when intersection is unsafe. • System tracks arrival time (or speed) of lead vehicle Speedometer • Speed monitor for lead vehicle. • Flashes red when near or far-side vehicle is speeding. Hybrid • Arrival time countdown for lead vehicle. • Prohibitive symbol relative to maneuvers based on near and far-side traffic conditions. Spit-Hybrid • Median position with logic for North • Left nearside position for North and South.

Baseline

Hazard Beacon

Speedometer

Hybrid

Split Hybrid

Benefit: Cost Analysis David Levinson u u u Identify relevant technologies: Review of literature Develop benefit cost framework Estimate lifespan of technology Estimate costs of technology Estimate benefits of countermeasures Lifecycle analysis Recommend countermeasures Analyze Inter-technology effects Determine performance metrics Develop cost: performance models Analyze synergies Optimize counter-measure combination

Predicting the Effect of IDS: A Statistical Model Gary Davis u Developed model to predict the Accident Reduction Effect of IDS deployment on Through -Stop rural expressway intersections in Minnesota u Adapted Accident Prediction Methods developed for FHWA's Interactive Highway Safety Design Module (IHSDM) u Presently Underway: Developing model to predict the Relative Risk to older drivers (not accident counts, as above) using similar methodology.

Predicting the Effect of IDS: A Statistical Model Gary Davis u Using 3 -year accident counts from the 198 4 -legged, 2 -way stopcontrolled intersections on Minnesota's rural expressways, v (1) Fit generalized linear model which relates expected accident count to major and minor ADTs, and no. of driveways within 250 ft of intersection. (Resulting model is similar to, but not identical with, the model developed in the IHSDM for two-lane highways. ) v (2) Used a hierarchical Bayes method to identify which intersections had atypically high accident counts. The intersection of USTH 52 & Goodhue CSAH 9 was one of these. v (3) Used estimates from part (2) to compute estimated Accident Reduction Factor as function of 1 -year's accident count after installation of IDS. v (4) Based on assessment of bounds of Accident Reduction Factor for the IDS, we will be able to predict accident reduction effects at other rural expressway intersection.

Previous Gap Acceptance Studies: Safe Gaps v. Left turn from a minor road – 8. 0 secs + 0. 5 secs for each additional lane to be crossed v. Right turn from a minor road – 7. 5 secs v. Crossing maneuver – 6. 5 secs for passenger cars, 8. 5 secs for single-unit trucks and 10. 5 secs for combination trucks; Add 0. 5 secs for each additional lane Source: National Cooperative Highway Research Program, Report 383, Intersection Sight Distance, National Academy Press, 1996. Basis of the Highway Design Manual for Older Drivers and Pedestrians, Publication No. FHWA-RD-01 -103, May, 2001, U. S. Department of Transportation. (See RECOMMENDATIONS: I. INTERSECTIONS (AT-GRADE), D. Design Element: Intersection Sight-Distance Requirement). See http: //www. tfhrc. gov/humanfac/01103/chp 1 rec. htm

Post-2005 Steps To FOT: u u u u Simulator is being used to evaluate relative merits of DII and to downselect the needed features of DII However, speed and gap size not perceived on road the same way as in simulator. Safe vs unsafe gaps: Used general guidelines from NCHRP 383 - for older drivers, use 8 sec instead of 7. 5 for left turn: 6. 5 sec for right turn, etc. Should conduct series of studies at isxn to model and differentiate needs between older and younger drivers, rather than use Hwy capacity/safety manual’s “recommended” values. What is the critical gap? For older drivers? For younger drivers? Need control study of old/young drivers on test intersection. Use Veh. DAQ/eye gaze tracking. Drivers take gaps confidently or not? Where to locate DII based on eye gaze study. Is DII intuitive? Will know state of vehicles on expressway and minor leg.

Post-2005 Steps To FOT: u u u Baseline will not need communications to/from vehicles. Will be able to test in MN plus 7 (? ) additional states. If FOT only evaluates DII, can then proceed immediately Cooperative Vehicle-Infrastructure systems: v Wireless communication to/from infrastructure v Vehicle data to RSU, then fused with other data to compute gaps v Driver (older, younger) and vehicle data to RSU, to determine safe vs unsafe gap v RSU to vehicle/driver DVI to provide isxn map; to inform driver v Nature and location of DVI in vehicle. Use eye-gaze measurement Pilot FOT v DII Infrastructure only; coop system; combined DII and DVI Large scale FOT v DII Infrastructure only; coop system; combined DII and DVI Develop Design Handbook/Warrants

At the Intersection

Two crashes since construction began in May. Right angle crash resulted in injuries (stretcher and ambulance)

MN Pooled Fund Project: Towards a Multi-State Consensus Minnesota is leading a state pooled fund project for rural intersection IDS, includes … MN, NV, NH, WI, MI, GA, IA, NC Multiple goals for state pooled fund: u Assistance/buy-in for DII design v Goal: nationally acceptable designs • Performance, Maintenance, Acceptability • Interoperability u Increased data collection capability v v Test intersections in participating states Regional vs. national driver behavior