U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000


Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

 
REPORT
This report is an archived publication and may contain dated technical, contact, and link information
Back to Publication List        
Publication Number:  FHWA-HRT-17-107    Date:  March 2018
Publication Number: FHWA-HRT-17-107
Date: March 2018

 

Identification of High Pedestrian Crash Locations

CHAPTER 4. INTERVIEWS ABOUT PEDESTRIAN COLLISION WARNING SYSTEMS

INTRODUCTION

The existing process for identifying unsafe locations for pedestrians is often based on accumulating pedestrian–motor vehicle crashes over several years until a high crash pattern emerges at specific locations. But what if engineers and planners did not have to wait until too many crashes happen to be able to identify these unsafe locations? What if engineers and planners had the ability to quickly (within weeks or months) identify where a lot of near crashes occur between pedestrians and motor vehicles, without waiting for years for high numbers of pedestrian crashes to occur?

Automated pedestrian collision warning systems in newer cars have the potential to dramatically improve the way engineers and planners identify unsafe locations for pedestrians. These warning systems use advanced sensors to detect when pedestrians are in proximity to a car’s trajectory and then provide a warning to the driver. In some collision warning systems, the car brakes may be automatically activated to avoid a collision with the pedestrian.

Conceptually, these near-crash collision warning events could be logged by a car’s computer system and then communicated to a centralized database. If these near-crash events are accumulated over all vehicles with pedestrian collision warning systems, there is the potential to have a citywide view of near crashes. If near crashes are an adequate surrogate for actual crashes, unsafe locations for pedestrians could be easily identified within weeks or months rather than years.

Research conducted by Leidos, Inc., for FHWA in 2014 and 2015 identified numerous vehicle-to-pedestrian (V2P) technologies that are capable of providing warnings to motor vehicle drivers or pedestrians of impending pedestrian–vehicle crashes.(64) The in-vehicle pedestrian collision warning systems described earlier are termed “unilateral pedestrian detection and driver notification” because the collision warning is provided only to the driver. This FHWA V2P effort resulted in the development of a Research Implementation Plan in May 2015.(65) However, none of the FHWA V2P documents presented the concept of using pedestrian collision warning events as surrogates for actual pedestrian crashes.

IN-VEHICLE SYSTEMS

Several new car models come equipped with collision avoidance systems that include pedestrian detection. These warning systems use advanced sensors to detect when pedestrians are in proximity to the car’s trajectory and then provide a warning to the driver. In some collision warning systems, the brakes may be automatically activated to avoid a collision with the pedestrian. Consider the following examples:

RESEARCH APPROACH

In task 3 of this project, the research team contacted three industry representatives to discuss the feasibility of logging warning events from pedestrian collision warning systems for use in pedestrian safety analyses. The following three companies were contacted:

The following section summarizes the findings from these three companies.

INTERVIEWS

Mobileye

Mobileye is a global leader in vision-based driver assistance and collision avoidance systems, and its machine vision products can be found in many common production car models.(66) Mobileye also works with a number of tier 1 partners that supply automotive components to other car manufacturers. They also have an aftermarket single-camera system suitable for trucks and passenger cars that can provide alerts and hot spots when connected to select devices.

In addition to its car-based collision avoidance systems, Mobileye has developed a pedestrian and bicyclist collision avoidance system for trucks, buses, and commercial vehicles called Mobileye Shield+™.(67) The Mobileye Shield+™ system uses similar machine vision technology to detect and track pedestrians and bicyclists in proximity to large vehicles. However, Mobileye Shield+™ differs from most auto-based collision avoidance systems in that it uses several strategically placed cameras to “see” in blindspots at the rear of these large vehicles. The Mobileye Shield+™ system has been installed by several transit authorities and other fleet vehicle operators in an effort to reduce pedestrian and bicyclist collisions.

These two types of collision avoidance systems—consumer-auto based and fleet-vehicle based—are addressed separately in this discussion. Fleet-vehicle-based collision warning systems represent the most feasible near-term implementation for logging and analyzing collision warning system events. Fleet vehicle operators have a strong vested interest in analyzing and understanding collision warning events in which their drivers are involved and in using this information to improve both fleet vehicle and roadway safety and operation. For consumer-auto-based systems, most consumers tend to be more concerned about their individual privacy and are hesitant to voluntarily provide data about their driving patterns, especially events such as collision warnings.

For Mobileye’s consumer-auto-based systems, it is not currently feasible for Mobileye to log collision warning events generated by its machine vision technology in consumer autos. Essentially, Mobileye is a parts supplier for automakers and other tier 1 parts suppliers and does not currently have a communication mechanism with new consumer autos equipped with advanced collision warning systems. If a consumer auto does have a communication mechanism (e.g., connected car), it could be with the automaker that sells the car or with an information services provider that offers location-based services (e.g., real-time navigation or travel time information). In these cases, the consumer who purchases the car must agree to terms and conditions that permit data to be retrieved from the onboard computer systems. However, these terms and conditions are often buried in long user agreements that few consumers actually read or fully understand (i.e., similar to lengthy software user agreements on desktop computers or smartphones). In summary, Mobileye is not currently involved in connected car communication with consumer autos equipped with its collision avoidance technology, thereby making it currently infeasible for Mobileye to log pedestrian collision warning events from consumer autos.

For Mobileye’s fleet-vehicle-based systems, it is feasible for Mobileye to log collision warning events. Mobileye has developed an analytics platform for summarizing these Mobileye Shield+™ collision warning events. The primary reason for differences of warning system data availability between consumer autos and fleet vehicles is that Mobileye sells the Mobileye Shield+™ system directly to the fleet vehicle operator and is, therefore, in the position to provide the communication mechanism needed to retrieve warning system event data from each equipped fleet vehicle. In fact, Mobileye markets the Mobileye Shield+™ system to include a “ full telematics system which tracks the vehicle and reports all warnings made by the Mobileye System to your fleet management system, providing fleet managers with valuable information about their drivers’ daily driving behavior.”(66)

Several transit agencies have already conducted pilot evaluations of the Mobileye Shield+™ system, including the Metropolitan Transit Authority in New York and King County Metro in Seattle, WA. As part of USDOT’s Smart City Challenge, Mobileye is planning to equip 300 transit buses in Columbus, OH, with the Mobileye Shield+™ system.

In partnership with another laboratory, the researchers also conducted a pilot evaluation of the Mobileye Shield+™ system on one of the busiest routes on the Texas A&M campus. The evaluation of the Mobileye Shield+™ system was conducted from January to March 2016. During this period, the equipped bus was in active service for 27 d and accumulated 41 pedestrian collision warnings (based on an estimated time to collision of 1.5 s). These pedestrian collision warnings were issued to the bus driver using both visual and audio alerts. These warnings and other sensor events were also communicated to a central repository and saved for historical analysis. This Mobileye Shield+™ evaluation illustrates the near-term feasibility of logging and analyzing collision warning system event data.

Since the exact location of each collision warning event was logged by the Mobileye system, the events could be easily mapped to show the frequency of these events. Mobileye had already created a hot spot locator that enabled a crash frequency map to be generated in a point-and-click Web browser interface ( figure 1 ). These hot spot locations were shared with the Texas A&M bus operators and dispatchers, who confirmed they were known locations where bus–pedestrian–bicyclist conflicts were common.

Screenshot. Mobileye interface for mapping warning event hot spot locations. This graphic shows a portion of the website that would demonstrate the locations of events of interest. For this graphic, a portion of the Texas A&M Campus is the background, and warning symbols mark the intersections where events occurred between January 1, 2017, and November 6, 2017. The total number of events is shown on the warning symbol with the number of events of 9 in the parking lot near Wellborn Road, 7 at University Drive and Tauber Street, 12 on Ross Street near Houston Street, 13 on Ross Street and Asbury Street, 5 at Ross Street and Ireland Street, 60 and 10 on the Ross segment between Ireland and Spence, 42 near the intersection of Ross Street and Spence Street, 6 near the intersection of parking lot and Ross Street, 6 at the intersection of Bizzell Street and Ross Street, 12 on Ross Street near bus stop, 6 near intersection of Bizzell Street and Polo Road, 10 at Bizzell Street and Lubock Street, 6 at intersection of Lubbock Street and parking lot entrance, 21 at intersection of Lubbock and Nagle Street, 18 at the intersection of Lubbock Street and Coke Street, 5 at the intersection of Coke Street and Routt Road, 18 at the intersection of Throckmorton Street and Routt Road, 17 at intersection of Routt Road and Houston Street, 34 near Rudder Fountain, and 8 near Simpson Drill Field. The graphic includes two boxes superimposed on the base map. The first box says “MobilEye Events for the Fleet,” “Start date: 01/01/2017,” “End date: 11/06/2017,” “Event types: 4 Selected.” This box also includes the option of engaging the following two filters: “Week-Days (Filtered by days)” and “Choose time range (Use time filter).” The second box states “MobilEye events mapping Total Number of Events-76469.” It provides two examples of events: 60 events for the address of “575 Ross St, College Station, TX 77840, USA,” and 42 events for the address of “710 Ross St, College Station, TX 77840, USA.”

Source: Evaluation of Mobileye Shield+TM.

Figure 1. Screenshot. Mobileye interface for mapping warning event hot spot locations.

This Mobileye system for fleet vehicles serves as a proof of concept for saving and analyzing pedestrian collision warning events. However, the limitation to fleet vehicles may restrict the representativeness of the data gathered. For transit systems, the warning event data are limited to the bus routes driven by equipped buses. For other fleet vehicles, the warning events may be unique to that type of fleet vehicle and not indicative of normal consumer cars and trucks.

Toyota

Toyota is a global auto manufacturer that produces several car models with its precollision system with pedestrian detection function.(68) According to the Toyota representative, Toyota does not have access to the warning system data generated by its precollision system. That is, Toyota does not currently have the communication mechanism in place to permit the routine logging of pedestrian collision warning events. In actual crash situations, Toyota Motor Sales investigators may retrieve data from the in-vehicle computer system, but this data retrieval is performed onsite on a case-by-case basis, and the data gathered in this situation are considered confidential.

The Toyota representative indicated that the Insurance Institute for Highway Safety’s Highway Loss Data Institute has gathered postcrash data from onboard collision warning systems (i.e., forward crash and lane departure) in an effort to estimate the value and benefits of collision warning systems.(69) However, this effort is not systematic across the entire equipped vehicle fleet (i.e., it includes only vehicles in crashes that have an insurance claim) and has not been performed for pedestrian collision warning systems yet due to their very low vehicle fleet penetration.

HERE

HERE is a global location-based data services company that also provides mapping technology used in advanced driver assistance systems. HERE gathers and systematizes location content (e.g., road networks, buildings, parks, points of interest, and traffic patterns) and then sells or licenses this location content to carmakers, various other businesses, and government entities.

Of particular interest in this discussion are the connected vehicle services that HERE currently provides (or plans to provide) to several carmakers. To provide these information services to consumer cars, HERE must maintain a communications mechanism that enables two-way data flow. That is, HERE first gathers various sensor data from the connected cars of several automaker brands; second, HERE anonymizes, aggregates, and analyzes the data to develop actionable information; and third, HERE communicates the actionable information to the relevant connected cars via the cellular network or short-range communications networks. For example, for a temporary road hazard where some drivers are using hard braking ( figure 2 ), HERE would gather these hard-braking events, and HERE’s location analytics platform would infer that something is causing an unexpected slowdown at that location. HERE would then communicate a targeted traffic disruption/slowdown warning to the automaker’s connected cars that are approaching the location.

Graphic. Simplified concept of HERE’s car-to-cloud data transmission specification. This graphic shows an illustration of cars on a level roadway with two horizontal curves where one car has run off the road following a puddle. Mountains are shown in the background along with one sign on a roadside signpost. The primary objective of the graphic is to illustrate the lines of communication. Communication lines are drawn between vehicles. They are also drawn between vehicles and the “Sensor Ingestion Interface Specification” bubble and between one of the vehicles and the “HERE Location Cloud” bubble. A line of communication is shown between the “Sensor Ingestion Interface Specification” bubble and the “HERE Location Cloud” bubble. A line of communication is also shown between the car and the “HERE Location Cloud” bubble with the text “analyzed data is sent to relevant vehicles in proximity.”

Source: http://360.here.com/2015/06/23/here-sensor-data-ingestion/, ©HERE.

Figure 2. Graphic. Simplified concept of HERE’s car-to-cloud data transmission specification.(70)

This simple example can be extended to all types of events that might occur in a connected car, including a pedestrian collision warning event. All that is required (from a technical perspective) is a real-time communications mechanism that can be used to gather event data from the onboard car computer system. Of course, from a business perspective, there must be some value in gathering and aggregating connected car data. New car models have extensive sensor systems that can generate terabytes of data per day, which requires extensive communications bandwidth and data disk storage. To maintain a sustainable business model, companies like HERE will have to make tough decisions about which connected car data are most valuable to collect and aggregate in a centralized database.

For confidentiality reasons, the HERE representative was not able to discuss specific company plans or actions about gathering pedestrian collision warning events and making them available for historical/analytic purposes. However, in the public realm, HERE has recently been working with the entire automotive industry on standardizing the information that is gathered from connected cars so that interoperability among various car models and information service providers is possible. In June 2015, HERE published a vehicle-to-cloud data specification that is intended to standardize the data that are transmitted from connected cars to centralized cloud-based databases.(71) This HERE data specification does address pedestrian and bicyclist detection from onboard car sensors but does not address warning system events. The following data attributes are included in version 2.0.2 of HERE’s data specification:

In June 2016, HERE submitted the design for a standardized data format for connected car sensor data to a European public/private partnership for intelligent transport systems, which has agreed to continue as an innovation platform to evolve the design into a standardized interface specification for broad use across the automotive industry. Recently, HERE announced it will be using this car-to-cloud data specification to gather sensor data from various car manufacturers.(72) Although the initial scale of this HERE activity could be modest (in terms of numbers of connected cars involved), it is considered the early steps on a path toward much greater levels of sensor data collection from connected cars.

 

 

Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000
Turner-Fairbank Highway Research Center | 6300 Georgetown Pike | McLean, VA | 22101