Parking crashes and Parking Assistance System Design: Evidence from Crash DataBases, the Literature, and Insurance Agent Interviews Paul Green

ABSTRACT

This paper (1) summarizes previous human factors/safety research on parking (8 studies, mostly over 20 years old), (2) provides statistics for 10,400 parking-related crashes in Michigan from 2000-2002, and (3) summarizes interviews with 6 insurance agents concerning parking crashes.

These sources indicate:

1. About 1/2 to 3/4 of parking crashes involve backing, often into another moving vehicle while emerging from a parking stall.

2. Eight-and-a-half foot-wide stalls had higher crash rates than wider stalls.

3. Most parallel parking crashes occur on major streets, not minor streets.

4. Lighting and driver impairment were minor factors in parking crashes.

INTRODUCTION

This paper describes background information for the Nissan Around View Monitor (AVM) project, a video-based system to assist with parking, with the specific purpose of helping to design that system to the U.S. market. The product goals – to make parking safer, easier, and more comfortable for drivers – served to focus this review. Particular emphasis for all evaluations was given to collecting data on collision geometry, vehicle type, driver age, and other information that might indicate the situations for which a system might be useful, what the driver needs to see, and who might use it.

The electronics content of motor vehicles, particularly computer-controlled systems, has grown considerably over the last few years. These systems control the engine and drive train, control and deploy air bags, present information to drivers, and perform many other applications. These systems can enhance not only the safety, but also the efficiency, usability, and convenience of motor vehicles.
Most research has focused on safety, with emphasis on the most severe and life-threatening crashes. In contrast, this research concerns parking-related crashes, which result in few deaths and injuries, but for which property damage is quite common. Parking is a difficult task drivers do not enjoy.
Accordingly, a significant number of devices have emerged in the market to make parking safer and easier. These include devices that use sonic sensors and provide audio feedback, usually in the form of a beep, when drivers approach an object. Other systems, such as those found on several current-model Infiniti vehicles, have rear-mounted cameras whose image is displayed on the instrument panel. These systems support backing maneuvers.
To aid Nissan in the development of a more sophisticated system, UMTRI conducted a number of experiments that involved collecting data on the size of parking spaces and how accurately people typically park (Cullinane, Smith, and Green, 2004), how well drivers parallel (Walls, Green, Gadgil, Amann, and Cullinane, 2004) and perpendicular park (Walls, Amann, Cullinane, Green, Gadgil, and Rubin, 2004) with camera systems that provide a 360 degree field of view, and the desired clearance around a parking vehicle (Gadgil and Green, 2005; Green, Gadgil, Walls, Amann, and Cullinane, 2004). All of those studies led to a set of design and evaluation guidelines that were summarized in a technical report (Rubin and Green, 2005). The current paper is an abridged version of the report written early in the project that examined the problem background (Smith, Green, and Jacob, 2004), namely the prior studies of parking, the relevant crash literature (in particular crashes in Michigan), and the experience of insurance agents with parking crashes. The author of this paper has taken the liberty of liberally reproducing material from that report (Smith, Green, and Jacob, 2004) without providing exact page citations.
More specifically, questions of interest in those studies included:
1. What previous human factors research has been done on parking lot, driveway, and related crashes?

2. How often do various types of crashes (by collision geometry, vehicle type, driver age, etc.) occur at low speeds (25 mi/hr or less), especially those involving parking or driveways?

3. What do insurance agents say are typical characteristics of parking lot and low-speed crashes?
The answers to these questions should help determine which parking tasks could benefit from a parking assistance system. These answers will also (1) guide camera placement, aim, and field of view, (2) possibly provide suggestions for the desired image resolution and in-vehicle display location, and (3) possibly help identify the types of vehicles for which such systems are likely to provide the greatest safety and convenience benefits, as well as sales potential.
The Literature on Parking

A search of the UMTRI Library using “parking” and related terms uncovered less than a dozen items in that 40-year-old collection. Searching the web using Google.com and the same terms yielded no additional articles. The articles found primarily concerned crash rates as a function of the parking configuration (angled, parallel, or straight-in). Other issues included parking space turnover, driver characteristics, and parking space width. A summary of those articles (parking crashes, backing behavior), listed in chronological order, follows.

Studies of Parking Crashes

Box (1981) examined private property crash reports for 1978 through 1980 from a single small city, Naperville, Illinois. Table 1 shows the number of crashes during that period in which a stationary vehicle was struck by a moving vehicle. Injuries were rare (1/525 crashes, driving in an aisle). Nearly 60 percent of parking-related crashes involved a stationary vehicle being damaged by a vehicle backing out of a parking space. (Vision is often limited in these situations, but could be supplemented by cameras). Crashes while pulling into a space constituted 21 percent of all crashes.

Table 1. Moving vehicle striking parked vehicle

(Box, 1981)

Operation	Category	Property damage only	Total	% of total
Entering a space	Pulling in	30	30	21
	Backing out	9	9	3
Leaving a space	Pulling out	13	13	5
	Backing out	150	150	59
Other	Backing (not to park)	25	25	10
	Driving in aisle	25	26	10
	Driving between parked cars	2	2	1
Subtotal		254	255	100
Unknown		270	270
Total		524	525

Note: Based on other summary tables and the text, the original article appeared to contain erroneous counts for crashes involving an injury. Table 1 has been corrected accordingly.
Table 2 shows the distribution of crashes involving a moving vehicle striking another moving vehicle. Similar to Table 1, the movement most associated with crashes was backing out of a parking space. Again, injuries were rare (15/245 crashes).
Table 3 shows the types of fixed objects struck by a moving vehicle during a parking-related maneuver. Buildings were struck more than twice as often as any other category of objects (44 percent of crashes). Given that buildings are typically stationary and easy to see, the issue may be one of judging clearance, not of vision per se. Keep in mind that crashes involving a vehicle striking a fixed object only represent 6 percent of all parking-related crashes and, furthermore, perhaps many of the parking spaces were near buildings when the data were collected in Naperville 24 years ago.

Table 3. Moving vehicle striking fixed object

(Box, 1981)

Category	Property damage only	Injury	Total	% of total
Light pole	6		6	11
Utility pole	2		2	4
Curb		1	1	2
Building	22	1	23	44
Tree	3	1	4	8
Fence	5		5	10
Other (signs, rocks, water)	9	1	10	19
Fall from fender		1	1	2
Total	47	5	52	100

Note: “Fall from fender” is not defined in the original report.

For a sample of parking lots where most of the crashes occurred, Table 4 shows crash frequency per space increased as parking turnover (3 undefined qualitative levels) increased. However, the differences in crash frequency as a function of stall angle are difficult to explain, 58 for 45 degrees, 80 for 60-75 degrees, and 19 for 90 degrees.
Table 4. Angle of stall relation to crashes/1000 spaces (Box, 1981)

	Turnover classes
Stall angle (degrees)	Low	Medium	High	Total
45-50	7.1	17.3	79.9	58.1
60-75	0.0	47.0	94.8	79.6
90	17.9	13.9	50.9	19.4
Total	16.7	18.3	79.4	36.4

As shown in Table 5, parking space width had no overall effect, except that the crash frequency was greater for the narrowest stall width (8.5 feet), though the interactions with turnover were considerable. Since 1981 when this study was conducted, the vehicle fleet in the U.S. has changed considerably, with large, wide vehicles (SUVs, light trucks, minivans) comprising a much larger fraction of the vehicle fleet. Hence, these results should be applied with caution.

Table 5. Width of stall relation to accident rate

(Box, 1981)

Stall width (feet)	Turnover classes			Total
	Low	Medium	High
8.5	11.8	95.2	28.0	53.1
9.0	17.8	17.6	71.6	35.1
9.5	16.0	3.5	118.5	30.6
10.0	25.6	18.1	69.2	39.5
Total	17.6	17.4	77.1	35.4

McCoy, Ramanujam, Moussavi, and Ballard (1990) examined crash rates for 1985 and 1986 on 491 urban sections of Nebraska state highways that could contain on-street parking. Only crashes with parked vehicles and those resulting from parking maneuvers were included.

Overall, parking maneuvers led to 37 percent of all non–intersection crashes, crashes examined in detail in Table 6. Those crashes occurred more often on 2-way, 2-lane roads than on major streets, and less for parallel than angle parking.
Table 6. Percentage of parking crashes as a function of parking and road type (McCoy, Ramanujam, Moussavi, and Ballard, 1990)

Type of parking		Major streets	2-Way, 2-lane streets
Painted	Parallel	33	46
	Low-angle	-	77
	High-angle	44	81
Unpainted	Parallel	21	39
	Angle	33	67

Readers should keep in mind that the various parking situations in Table 6 do not occur equally often. Table 7 shows non-intersection and parking crash rates respectively, that is, exposure-adjusted data. Interestingly, the conclusions are the same -- crash rates are greater for 2-way, 2-lane roads than major streets and that crash frequency (per space) is less for parallel parking than other types of parking. However, when adjusted for exposure, painting lines reduces crashes (in terms of frequency per stall).

Table 7. Non-intersection and parking crash rates

(McCoy, Ramanujam, Moussavi, and Ballard, 1990)

			Non-intersection		Parking
Item	Markings	Type	Major streets	2-Way, 2-lane streets	Major streets	2-Way, 2-lane streets
Crashes (million vehicle -miles)	Painted	Parallel	1.65	1.83	0.550	0.848
		Low-angle	-	3.38	-	2.60
		High-angle	1.20	3.59	0.533	2.91
	Unpainted	Parallel	1.32	0.674	0.284	0.264
		Angle	1.57	1.67	0.524	1.11
Crashes (10 billion vehicle-mile-hours/stall)	Painted	Parallel	6.50	6.58	2.17	3.05
		Low-angle	-	9.59	-	7.38
		High-angle	7.19	12.9	3.19	10.5
	Unpainted	Parallel	7.67	5.44	1.65	2.13
		Angle	13.19	12.10	4.40	8.04

An undated study by Seburn, probably from the mid to late 1960s, aggregated data from 32 cities in 17 states and the District of Columbia. Some 46 percent of crashes involved a vehicle traveling straight ahead on a roadway and colliding with a parking vehicle or one performing a parking-related maneuver (Table 7). As shown in Table 8, nearly 65 percent of parking crashes involved vehicles backing into or pulling towards a curb, and are likely associated with parallel parking.
Table 7. Number of crashes by vehicle operation

Vehicle motion	Number	%
Traveling straight ahead	1639	46.0
Entering a parking space	1211	34.0
Leaving a parking space	355	10.0
Turning (driveway or intersection)	352	10.0
Total	3557	100.0

Table 8. Number of crashes by parking operation

Parking maneuver	Number	%
Backing into curb	414	26.4
Pulling to curb-forward	609	38.9
Pulling from curb-forward	378	24.1
Stopped-not at curb	83	5.3
Unknown	82	5.3
Total	1566	100.0

Figure 1 shows the percentage of parking crashes by driver age groups, along with similar data for all crashes. The percentage of drivers in each category is not provided. There is a tendency for drivers under age 25 to be relatively more involved in parking crashes, which suggest they could benefit most from parking assistance devices.

Figure 1. Age of drivers involved in parking crashes (Saburn, undated)

Finally, Saburn (undated) concluded that lighting (or possibly fatigue) did not affect parking, even though 46 percent of all driving and 43.5 percent of all crashes occur at night.
Humphreys, Box, Sullivan, and Wheeler (1978) reported an analysis of street and crash data from 10 cities representing 5 states. Of the 4,800 crashes examined, nearly 3,600 were midblock crashes or intersection crashes in which parking was deemed to be a factor. Parking was a factor in 54 percent of midblock and intersection crashes, and 85 percent of those crashes involved damage to the vehicles but no injuries (Table 9).

Table 9. Crashes by severity, street class, and parking involvement

	Property damage only		Injury
Street class	Parking involved	Other	Parking involved	Other	Total	Percent parking involved
Local	396	133	37	37	603	72
Collector	150	60	8	15	233	68
Major	1229	1094	112	323	2758	49
Total	1775	1287	157	375	3594	54

Table 10 shows 2-way major street midblock crashes by type and parking involvement. Parking was involved in all crashes where a vehicle hit a parked car, in 36.8 percent of crashes involving a pedestrian, and in about 5 percent of crashes where a vehicle struck another vehicle or a fixed object. Parking was involved in 43 percent of crashes on 2-way major streets, and a similar analysis for 1-way streets showed that parking was involved in 60 percent of crashes.

Table 10. Two-way major street midblock crashes by type and parking involvement

Type of parking involvement	On street	At driveway	Fixed object	Parked car	Pedestrian	Bicycle	Misc	Total
No involvement	518	315	154	0	24	9	16	1,036
Open door	1	0	0	59	0	1	0	61
Entering space	6	1	3	73	2	0	0	85
Sight restricted	0	2	1	0	8	0	0	11
Stationary	11	5	1	357	2	0	0	376
Leaving space	8	6	4	222	2	0	0	242
Parking subtotal	26	14	9	711	14	1	0	775
Total	544	329	163	711	38	10	16	1,811
Percent parking involved	4.7	4.2	5.5	100	36.8	10	0	43

Box (2001) reviewed published and unpublished curb parking studies from the 1970s through 2001 with the purpose of summarizing studies of parallel and angle parking, with an emphasis on angle parking. Several of those studies have been previously discussed. Of those not reviewed, several concerned small towns and a limited number of crashes. In general, they suggested that crash rates were lower for parallel parking than for angle parking.
Studies of Parking Maneuvers

The predominance of backing-related crashes led to a focus on that maneuver. Harpster, Huey, and Lerner (1996) had 9 elderly and 12 young drivers back their own vehicles on public roads in real-world driving conditions. They are listed as a footnote to Table 11. That table shows that drivers glanced over their right shoulders while backing more than 50 percent of the time, which was 4 times greater than how often they looked over their left shoulders. Almost no glances were made to the dash during the backing tasks.

Table 11. Glance direction for each task (Harpster, Huey, and Lerner, 1996)

		Task
Glance direction	Total	1	2	3	4	5	6	7	8
Forward	10.6	28.7	16.5	0.6	1.8	18.1	3.0	14.1	1.8
Dash	0.0	0.1	0.0	0.0	0.0	0.0	0.0	0.1	0.0
Driver’s mirror	8.2	6.0	3.6	8.4	7.8	14.7	11.6	2.8	11.0
Rear mirror	4.5	4.7	7.4	4.0	1.7	4.5	2.3	6.3	5.1
Right mirror	9.2	7.1	17.8	2.7	3.7	13.4	2.7	19.3	7.2
Left window	2.3	1.5	0.3	0.2	0.2	4.9	1.3	0.0	0.0
Right window	1.0	3.0	3.6	0.1	0.3	5.1	1.2	4.6	0.4
Gear shift	1.0	3.2	-0.7	0.0	0.4	2.1	0.3	1.1	0.1
Left shoulder	12.5	21.6	4.2	27.3	6.0	14.0	14.4	5.2	7.2
Right shoulder	50.9	23.6	47.0	56.2	78.2	24.7	62.7	47.2	67.3

The 8 tasks included: (1) backing out of an angle slot in a parking lot, (2) parallel parking against a curb with vehicles in front, (3) extended curve backing to a stop point (location 1 of 2), (4) backing to a wall, (5) backing out of a perpendicular slot in a parking lot, (6) backing into a perpendicular parking slot, (7) parallel parking against a curb with vehicles in rear, and (8) extended curve backing to a stop point (location 2 of 2).

Given the restricted neck mobility of older drivers, there were major age differences in glance behavior (Table 12), with older drivers making greater use of mirrors and looking less over their right shoulder.
Table 12. Glance direction as a function of age across all tasks (Harpster, Huey, and Lerner, 1996)

Glance direction	Young	Elderly
Forward	9.9	11.8
Dash	0.0	0.0
Driver’s mirror	4.3	15.0*
Rear mirror	3.3	7.1*
Right mirror	7.7	12.1
Left window	0.7	1.5
Right window	2.1	2.1
Gear shift	0.4	1.8*
Left shoulder	12.8	12.3
Right shoulder	59.9	37.4*

Interestingly, as the drivers approached an object, they decelerated such that time to collision (TTC) remained relatively constant (Table 13). Although there were speed differences between younger and older drivers, there were no differences in minimum TTC.

Table 13. Minimum time to collision (MTTC) for all participants (s) (Harpster, Huey, and Lerner, 1996)

Task	Mean	Min	10th %	Max
2. Parallel	3.4	1.0	1.3	6.3
4. Back to wall	2.4	1.1	1.5	3.9
6. Back in perpendicular	3.0	1.7	1.9	4.3
7. Parallel	3.7	2.0	2.1	6.3

The author was able to find only one study concerning parallel parking performance. Literature on this topic is extremely limited, even though it is a very common maneuver. As part of an experiment on steering effort levels that involved driving on city streets, an oval test track, and a slalom, Green, Gillespie, Reifeis, Wei-Haas, and Ottens (1984) had 43 Ford employees parallel park a 1984 Ford Thunderbird and a 1984 Ford LTD sedan. The number of vehicle movements was recorded but not reported, as the experiment concerned desired steering effort levels. One noteworthy tendency the author recalls was for a few subjects to continue moving in a particular direction until they hit something, especially when backing. Thus, in a parallel parking study, damage to the test vehicle or adjacent vehicles is expected.

Summary of the Human Factors Literature

Thus, as a whole, these studies suggest that depending on the study, parking crashes represent between 1/3 and 1/2 of all police-reported crashes, and that backing out of an angle or perpendicular space is the most common scenario. For those parking maneuvers, drivers have more crashes with 8.5-feet wide stalls than with wider stalls. However, readers should keep in mind that many of the studies are from smaller cities. Further, the U.S. vehicle fleet has changed since many of these studies were conducted, with light trucks, SUVs, and minivans now being predominant. Some of these vehicles can be harder to see through, are much taller and wider, and have a much higher driver eye position. Furthermore, there is a dearth of studies from urban areas.

Only one study could be located with data describing driver-backing behavior. It showed that older drivers had different scanning patterns than younger drivers (fewer direct looks to the right side but more use of mirrors) and that backing occurred with a constant TTC of about 3 seconds.

Parking Crashes in Michigan

All police-reported crashes in the state of Michigan from 2000, 2001, and 2002 were examined. In Michigan, a traffic crash is defined as an incident involving a motor vehicle, in transport, on a roadway, that resulted in death, injury, or property damage of $400 or greater. Michigan data were selected because they are reasonably typical of the U.S.; contain a mixture of urban, suburban, and rural settings; and were reasonably accessible.

Of the 1.2 million crashes in the database, about 10,000 or 1 percent were coded as entering or leaving parking. However, the actual number of parking-related crashes is much larger as police-reported crashes mostly concern public roads but many parking-related crashes occur on private property, and therefore are often not in the database. As noted in Table 14, the number of crashes declined over the 3 years examined, with crashes involving leaving a space being 2.6 times more likely than crashes entering a space.

Table 14. Summary of crashes associated with entering and leaving parking

Crash type	Item	Year			Total
		2000	2001	2002
Park	Enter	1058	978	858	2894
	Leave	2818	2542	2207	7567
	Total	3876	3520	3065	10461
All	# Fatal	1237	1206	1175	3618
	# Injury	87043	80922	80567	248532
	Total	424852	400813	395515	1221180

Directory: ~driving -> publications
publications -> Methods for assessing the safety and usability of in-vehicle systems: Lessons from umtri driver interface research