The Light Vehicle - Heavy Vehicle Interaction Problem

In 2002, 434,000 large trucks (gross weight > 10,000 lbs) were involved in vehicle crashes; 4,542 of these crashes resulted in fatalities. In these crashes, 4,897 people died and an additional 210,000 were injured. Though accounting for 4% of all registered vehicles in 2002, large trucks represented 8% of all vehicles involved in fatal crashes (National Highway Traffic Safety Administration, NHTSA, 2003). However, truck drivers have lower non-fatal crash rates per million vehicle miles traveled than light vehicles (NHTSA, 2003). Nonetheless, light vehicles are extremely vulnerable when they interact with trucks because trucks often weigh 20-30 times as much as light vehicles (Insurance Institute for Highway Safety, 2002), and trucks take 20-40% farther to stop than light vehicles (NHTSA, 1987). A better understanding of LV-HV interactions is needed to develop future interventions and countermeasures directed at mitigating the problem. The data from the 100-Car Study (Dingus et al., 2004) were used in the current project to assess the LV-HV interaction problem from the LV drivers' perspective.

Project Goals

There were four primary goals in the current effort:

1. Gain a better understanding of LV-HV interactions on the nation's roadways.
2. Continue to develop the classification scheme and corresponding Contributing Factors list for LV-HV interactions used in Hanowski, Keisler, and Wierwille (2004) and use the terminology and methodology described in the LTCCS (Thieriez, Radja, and Toth, 2002).
3. Compare the current data to the data obtained in the Hanowski, Keisler, and Wierwille (2004) study for a more complete picture of the LV-HV interaction problem.
4. Provide background information that would serve as a necessary prerequisite to the development of countermeasures for LV-HV interactions.

Method

One hundred and nine participants who commute to and from the Washington, DC metro area were recruited as drivers in the 100-Car Study. The 109 participants ranged in age from 18 to over 55 years (43 female, 66 male). One hundred LVs were instrumented for this study; 80 vehicles were owned by the participants, while 20 were leased vehicles from VTTI.

The data used in the current effort consisted of video recordings of critical incidents. Five video cameras were used in the video recording system: (1) a forward-looking camera that captured the forward roadway scene, traffic situation, and possible incidents; (2) a driver's face camera that was used to record facial expressions, eyelid closure, glance position, and head turns; (3) a right-side camera that was mounted on the A-pillar of the passenger side and faced outward; (4) a dome camera that was mounted from inside the vehicle and faced over the drivers shoulder towards the steering wheel, hands, and feet; and (5) a rear camera that was intended to capture the situation behind the vehicle. Infrared lighting was used to illuminate the vehicle cab so that the driver's face as well as their hands could still be viewed by the camera during nighttime driving.

The video continuously recorded while the ignition was on, thereby allowing laboratory review and selection of the video without losses of any kind. The videotaped episodes/incidents were selected and keyed to digitally-recorded data.

Of interest in the data set were "critical incidents" defined as unexpected events resulting in a close call or requiring fast action (evasive maneuver) on the part of a driver to avoid a crash. Critical incidents were detected by one of three methods. The first method involved flagging events where the car sensors exceeded a specified value (e.g., brake response of >0.6 g). The second incident flagging method occurred when the driver pressed an incident pushbutton located on the dashboard (i.e., drivers were instructed to depress a button on the dashboard if they witnessed an incident or were involved in an incident). The third method of detecting incidents was through analysts' judgments when reviewing the video. Only those events that involved a LV-HV interaction are described in the current analyses.

Result Highlights

Light Vehicle-Heavy Vehicle Interaction Data Set

The 100-Car Study captured 9,125 incidents, which were divided into four categories: (1) LV-LV Interactions; (2) LV-HV Interactions; (3) Single Vehicle Conflicts; and (4) Other. Of the 9,125 events, 246 (2.7%) involved a LV-HV interaction.

Incident Types

With the 246 LV-HV interactions recorded in the data set, the next step in the analysis was to determine the vehicles' actions for each incident. To this end, the video and relevant data for each incident were carefully reviewed and then classified according to "Incident Type." Twenty-seven different Incident Types were identified in the data set.

The most frequent Incident Type involving a LV-HV interaction was Late Braking for Stopped/Stopping Traffic. Across all 246 incidents, this particular Incident Type occurred 66 times and accounted for 26.8% of the incidents captured. The majority of the incidents (48.8%) involved one of two different Incident Types: Late Braking for Stopped/Stopping Traffic, and Lane Change Without Sufficient Gap.

Descriptive statistics for the Incident Types were also calculated for incidents as a function of the at-fault driver. The at-fault driver is the driver that was assessed, by the analyst, to have been responsible for causing the event. Of the 246 LV-HV interaction incidents recorded, 138 (56%) were judged to have been the fault of the LV driver, while 79 (32%) were attributed to the HV driver. For the remaining 29 incidents (12%), it was unclear which vehicle driver was at-fault. By removing the "unknown" cases from the LV-HV driver at-fault analyses, it was found that the LV driver was at-fault in 64% (138/217) of the LV-HV interaction incidents, while the HV driver was at fault in 36% (79/217) of the incidents.

The most frequent Incident Type for HV driver at-fault incidents was Lane Change Without Sufficient Gap (26.6%), followed by Lateral Deviation of Through Vehicle (21.5%), and Left Turn Without Clearance (13.9%). The most frequent Incident Type for LV driver at-fault incidents was Late Braking for Stopped/Stopping Traffic (41.3%), followed by Lane Change Without Sufficient Gap (21.7%), and Aborted Lane Change (8%). The most frequent Incident Type for Unknown at-fault incidents was Late Braking for Stopped/Stopping Traffic (27.6%), followed by Conflict With Oncoming Traffic (13.8%), Lane Change Without Sufficient Gap (10.3%), and Unable to Determine (10.3%).

Primary Maneuvers, Secondary Maneuvers, and Conflict Types

After each of the 246 incidents was classified by Incident Type, the next step in the analysis was to identify the "Primary Maneuvers" and "Secondary Maneuvers" involved in each incident. The Primary Maneuver refers to the maneuver of the driver who initiated the incident (not necessarily at-fault). Across the 246 interaction incidents, 19 different Primary Maneuvers were identified. The most frequent Primary Maneuvers were Braking (22.7%), Changing Lanes (21.1%), and Stopped (15%). These three Primary Maneuvers represented 58.9% of the recorded incidents.

In addition to identifying the Primary Maneuver for each incident, the Secondary Maneuver, or the maneuver of the responding driver (i.e., driver of the second vehicle involved in the interaction), was also classified. Considering the maneuvers of both vehicles involved in the incident, a clear picture of the conflict, or Conflict Type, could be determined. A total of 66 different Conflict Types (i.e., Primary Maneuver and Secondary Maneuver combinations) were identified.

Contributing Factors

Just as the Incident Types describe the action or what happened during an incident, Contributing Factors provide likely reasons why an incident occurred. For each incident that was analyzed, a number of Contributing Factors were identified. It should be noted that the Contributing Factor categories were taken from Wierwille, Kieliszewski, Hanowski, Keisler, and Olsen (2000) and from the GES Physical Impairments screen (USDOT/NHTSA, 2003, p. 434). Due to the methodology used, where the data collection equipment was only instrumented in the LV, the Contributing Factor was based solely on the behaviors of the LV driver. Without cameras inside the HV there was no way to determine, with any degree of certainty, the behavior(s) of the HV driver. Even when the HV driver was judged to have been at-fault, the behaviors of the LV driver were identified. Put another way, for the events that were caused by the LV driver, the analyses considered the LV driver's behaviors that may have contributed to the event. For events where the HV-driver was at-fault, the analysis also considered the LV driver's behaviors. However, the consideration is for the LV driver behaviors that occurred as the driver reacted to the HV driver's actions.

Note that multiple factors could be assessed for each individual event (as such, the percentages for the factors total more than 100%). Across all 246 incidents, the most frequent Contributing Factor was Driving Techniques (49.5%), followed by Unknown (24%) and Distracted (18.7%). The most frequent Contributing Factor for HV driver at-fault incidents was Unknown (68.4%), followed by Driving Techniques (15.2%), and Distracted (11.4%). The most frequent Contributing Factor for LV driver at-fault incidents was Driving Techniques (70.3%), followed by Distracted (22.5%) and Aggressive Driving (22.5%). The most frequent Contributing Factor for Unknown driver at-fault incidents was Driving Techniques (48.3%), followed by Distracted (20.7%), Roadway Alignment (10.3%), and Unknown (10.3%).

Driver Distraction

A substantial number of the LV-HV incidents had Distraction listed as a Contributing Factor. Again, as indicated above, the incidents where Driver Distraction was mentioned refer to the behavior of the LV driver. The Distraction Contributing Factor was sub-divided into more distinct categories. See Table 22 for a listing of the frequency, percentage, and rank ordering for each sub-category in the Distraction Contributing Factor. The most frequent sub-category for the Distraction Contributing Factor was Talking/Listening on Cell Phone (21.7%), followed by Passenger in Adjacent Seat (13%), and Dialing Hand-Held Phone (8.7%).

Classification strategy used in the LTCCS

Accident Type

Each of the 246 LV-HV interactions were grouped by Accident Type based on the methodology used in the LTCCS (Thieriez, Radja, and Toth, 2002). Note that there was only one LV-HV crash recorded in the 100-Car Study. Therefore, using the Accident Types from the LTCCS does not reflect an absolute match, but rather a relative match. However, to facilitate future data comparisons with the near-crash data collected in the current study with other studies using the LTCCS, each of the 246 LV-HV interactions were coded using the LTCCS classification scheme. Because only one crash occurred, the closest match with respect to Accident Types was recorded for each incident.

Overall, the most frequent Accident Types were Scenarios 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (19.9%); 20/21: Same Trafficway/Same Direction: Rear-End: Approaches Stopped Vehicle (16.3%); and 28/29: Same Trafficway/Same Direction: Rear-End: Approaches Decelerating Vehicle (13%). These three Accident Types represented 49.2% of the Accident Types for all LV-HV incidents.

The Accident Types for HV and LV driver at-fault incidents differed. The most prevalent Accident Types for HV driver at-fault incidents were Scenarios 44/45: Same Trafficway/Same Direction: Sideswipe Angle: In Blind Spot (27.8%), 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (15.2%), and 25/25: Same Trafficway/Same Direction: Rear-End: Approaches Slower Constant Speed Vehicle (8.9%). These three Accident Types accounted for 51.9% of the HV driver at-fault incidents. The most prevalent Accident Types for LV driver at-fault incidents were Scenarios 20/21: Same Trafficway/Same Direction: Rear-End: Approaches Stopped Vehicle (26.8%); 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (22.5%); and 28/29: Same Trafficway/Same Direction: Rear-End: Approaches Decelerating Vehicle (17.4%). These three Accident Types accounted for 66.7% at the LV driver at-fault incidents.

The most prevalent Accident Type for HV driver at-fault incidents involved a Sideswipe Angle. By summing all the HV driver at-fault Accident Types that involved a Sideswipe Angle, it was found that 41.8% of the HV driver at-fault incidents were coded with this Accident Type. Conversely, the most prevalent Accident Type for LV driver at-fault incidents involved a Rear-End Approach. By summing all LV driver at-fault Accident Types that involved a Rear-End approach, it was found that 55.1% of the LV driver at-fault incidents were coded with this Accident Type.

Critical Reason for the Critical Event

To be consistent with the LTCCS (Thieriez, Radja, and Toth, 2002), the LV driver at-fault incidents were coded with a Critical Reason for the incident. The Critical Reason for the incident was considered the primary reason for why the incident occurred. More than one Critical Reason could be coded for each incident (10 of the recorded incidents were coded with two Critical Reasons). Only the LV driver at-fault incidents were coded with a Critical Reason because those vehicles were equipped with video of the driver. For the HV driver at-fault incidents, it was not possible to determine with any certainty what the driver was doing that contributed to the event; therefore, all HV driver at-fault incidents were coded as "Unknown reason for the critical event."

Overall, the most frequent Critical Reasons for LV driver at-fault incidents were Aggressive Driving Behavior (24.6%), Too Fast for Conditions (15.2%), and Misjudgment of Gap (13.8%). There were other interesting trends. Sixty-four of the 138 LV at-fault incidents (46.4%) were coded with at least one Critical Reason that was a risky driving behavior (i.e., Aggressive Driving Behavior, Too Fast for Conditions, Following too Closely, and Illegal Maneuver), while 22.5% of the LV driver at-fault incidents involved some type of awareness variable (i.e., Internal Distraction, Inattention, External Distraction).

Comparisons using the 100-Car Data, the Local Short Haul Data, and the Sleeper Berth Data

The current study builds on a previous project that classified critical incidents (crashes and near-crashes) recorded in two fatigue study with Local/Short Haul (L/SH) drivers and Sleeper Berth (SB) drivers (Hanowski, Keisler, and Wierwille, 2004). The two truck studies involved instrumentation in the truck and recorded events from the HV driver's perspective. In contrast, the current study recorded events as they unfolded from the LV driver's perspective. The events from the three studies were combined and the classifications were compared.

A total of 142 LV-HV interactions were identified in the L/SH study. Of these, 117 (82.4%) incidents were judged to be the fault of the LV driver, while the remaining 25 (17.6%) incidents were the fault of the HV driver. In the SB study, a total of 68 LV-HV interactions were identified. Of these, 47 (69.1%) were assessed to have been the fault of the LV driver, while the remaining 21 (38.9%) were the fault of the HV driver. Of the 246 LV-HV interaction incidents recorded in the current study, 138 (56%) were judged to have been the fault of the LV driver, while 79 (32%) were attributed to the HV driver. For the remaining 29 incidents (12%), it was unclear which vehicle driver was at-fault. Considering the combined data, these three studies consistently show that LV drivers were judged to be responsible for the majority of LV-HV interactions. Of the 427 LV-HV incidents identified across the three studies (excluding the 29 Unknown at-fault incidents in the current study), 302 (70.7%) were judged to have been the fault of the LV driver, while the remaining 125 (29.3%) were the fault of the HV driver (a 2.4:1 ratio).

There were a number of interesting findings from the comparisons between the 100-Car, SB, and L/SH studies. Comparisons were conducted with respect to the Incident Type, Primary Maneuver, and Contributing Factor. The Incident Type comparison indicated that Lane Change Without Sufficient Gap was the most frequent Incident Type across all three studies. A breakdown of incidents as a function of the at-fault driver showed that Lane Change Without Sufficient Gap incidents were primarily attributed to LV drivers. Critical incidents that involved a LV driver changing lanes in front of an HV, leaving the HV driver with very little headway between vehicles, were a common Incident Type that was captured in all three studies.

While the Incident Types for the LV driver at-fault incidents shared some similarities across the three studies, the Incident Types for the HV driver at-fault incident were more varied across the studies. In the 100-Car Study, 48.1% of the HV driver at-fault Incident Types included Lane Change Without Sufficient Gap or Lateral Deviation of Through Traffic. In the SB study, 71.4% of the HV driver at-fault incidents included Late Braking for Stopped/Stopping Traffic and Following Too Closely. In the L/SH study, 48% of HV at-fault incidents included Roadway Entrance Without Clearance, Wide Turn Into Adjacent Lane, or Late Braking for Stopped/Stopping Traffic.

One possible explanation for these differences was the predominant Road Type traveled in each study as well as the predominant trucking operations in the SB and L/SH studies. However, it could be argued that the HVs in the 100-Car Study represent a more diverse population of HVs since they were not limited to L/SH and SB trucks. In fact, 25 different HV types were identified as being involved in LV-HV interactions in the 100-Car Study. Thus, it is likely the results for at-fault HV drivers in the 100-Car Study might be more representative of HV drivers in general, while the results for HV drivers in the SB and L/SH studies may provide greater insight for these specific operations.

Recall that the Contributing Factors category describes why the incident occurred. The most frequent Contributing Factor across the three studies was Driving Techniques. A breakdown of incidents as a function of the at-fault driver showed that Driving Techniques were primarily attributed to HV drivers. Thus, when the Contributing Factor was known, Driving Techniques was the most frequent Contributing Factor for HV driver at-fault incidents in each of the studies. Similarly, the most frequent Contributing Factors for LV driver at-fault incidents across the three studies were Driving Techniques and Aggressive Driving. These two Contributing Factors accounted for 69.7% of the LV driver at-fault incidents across the three studies. However, a large proportion of the LV driver at-fault incidents in the 100-Car Study involved the Distracted Contributing Factor. In fact, the only time a LV driver at-fault incident was coded with the Distracted Contributing Factors was in the 100-Car Study. This is due to the methodological approach used in that only the LVs in the 100-Car Study were instrumented (thereby allowing analysis of the LV drivers' behaviors while driving) while the LVs in both the SB and L/SH studies were not instrumented.

Conclusions

The analyses that were conducted with the LV-HV interactions captured in the 100-Car Study provide convincing evidence to support the contention that LV-HV interactions are a serious problem. While the 100-Car Study captured 9,125 critical incidents, only 246 LV-HV interactions (2.7%) were identified. While 2.7% may appear to represent a small proportion of the overall critical incident picture, it should be noted that LV-HV interactions have the potential to become serious, and even fatal due to the tremendous difference in weight between an HV and LV.

There are six key findings that stem from the analyses conducted on the interactions between HVs and LVs. First, of the 246 interactions that were analyzed, 138 (56.1%) were assessed to have been the fault of the LV driver. HV drivers were at-fault in 79 (32.1%) of the incidents, while in the remaining 29 (11.8%) incidents, it was unknown if the HV or LV driver was at-fault. Excluding the incidents where it was unknown if the HV or LV driver was at-fault, 63.6% and 36.4% of the incidents were the fault of the LV and HV drivers, respectively. Thus, LV drivers were judged to have been responsible for a substantial proportion of the LV-HV interactions. These findings support what the drivers in the Hanowski et al. (1998) focus groups reported about LVs being their most important safety concern. Further, the results are similar to prior published studies that used a crash database approach to assess LV-HV interactions (cf. Blower, 1998; Stuster, 1999; Wang, Knipling, and Blincoe, 1999). Based on these findings, it is suggested that focusing on the LV driver, and their errors, may provide the largest area of opportunity for reducing such events.

The second important finding from these analyses was in regard to the different Incident Types that were frequent among HV and LV drivers. For LV driver at-fault incidents, the most frequent Incident Types were: Late Braking for Stopped/Stopping Traffic (41.3%), Lane Change Without Sufficient Gap (21.7%), and Aborted Lane Change (8%). These particular Incident Types are indicative of at-risk driving behaviors. Once again, the objective data support the sentiment of the L/SH drivers in the Hanowski et al. (1998) focus group who indicated that during their daily travel, they were often "cut-off" by LV drivers. In addition, the data supports the results from the L/SH on-road study (Hanowski, Keisler, and Wierwille, 2004) where the most prevalent Incident Type for LV driver at-fault incident was Lane Change Without Sufficient Gap (accounting for 24.8% of the LV driver at-fault incidents in the L/SH study). In contrast, the most frequent Incident Types for HV at-fault drivers were: Lane Change Without Sufficient Gap (26.6%), Lateral Deviation of Through Vehicle (21.5%), and Left Turn Without Clearance (13.9%).

The third finding is the difference in the Primary Maneuvers for HV and LV drivers. The most frequent Primary Maneuvers for LV driver at-fault incidents were: Braking (32.6%), Stopped (21.7%), and Changing Lanes (16.7%). The two most frequent Primary Maneuvers for LV driver at-fault incidents involved assumed difficulties on the part of the LV driver decelerating or stopping. In contrast, the most frequent Primary Maneuvers for HV driver at-fault incidents were: Changing Lanes (32.9%), Crosses Over Lane Line (20.3%), and Left Turn (15.2%). The two most frequent Primary Maneuvers for HV driver at-fault incidents involved difficulties changing or crossing over the lane line while the vehicle was in motion. These results make intuitive sense because HV drivers have limited visibility and deal with blind spots thereby making lane changes difficult in traffic.

The fourth important finding is related to the Contributing Factors that were most frequent with LV and HV drivers. For LV drivers, the most frequent Contributing Factors for at-fault incidents were: Driving Techniques (70.3%), Distracted (22.5%), and Aggressive Driving (22.5%). The most frequent Contributing Factors for HV driver at-fault incidents were: Unknown (68.4%), Driving Techniques (15.2%), and Distracted (11.4%). The large number of Unknown Contributing Factors for HV driver at-fault incidents is indicative of the methodology used to code these events. Because the HV did not have any video cameras, the Contributing Factor was coded with respect to the behaviors of the LV driver. As the LV driver was not responsible for the incident, it was unlikely they would be coded with a Contributing Factor, thus the high frequency of Unknown Contributing Factors. Further, the methodology used to code the Contributing Factors also explains the similarities between LVs and HVs (i.e., they were all coded with respect to the LV driver, and therefore, they should be similar).

The fifth noteworthy finding from the current research involves the Accident Types (using the LTCCS approach and terminology) that were most prevalent for LV and HV drivers. The most prevalent Accident Types for LV driver at-fault incidents were: Scenarios 20/21: Same Trafficway/Same Direction: Rear End: Approaches Stopped Vehicle (26.8%); 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (22.5%); and 28/29: Same Trafficway/Same Direction: Rear End: Approaches Decelerating Vehicle (17.4%). Approximately 55% of the LV driver at-fault incidents involved a Rear End approach. These Accident Types also support the findings from the analysis of the most prevalent Primary Maneuvers for LV driver at-fault incidents: decelerating or stopped. Conversely, the most prevalent Accident Types for HV driver at-fault incidents were: Scenarios 44/45: Same Trafficway/Same Direction: Forward Impact: Sideswipe Angle: In Blind Spot (27.7%); 38/39: Same Trafficway/Same Direction: Forward Impact: Avoid Collision with Vehicle (15.2%); and 25/25: Same Trafficway/Same Direction: Rear End: Approaches Constant Speed Vehicle (8.9%). Approximately 42% of the HV driver at-fault incidents involved a Sideswipe Angle. These Accident Types also support the findings from the most prevalent Primary Maneuvers for HV driver at-fault incidents: changing lanes and crossing the lane line.

The sixth noteworthy finding from the current research reflects some of the similarities and differences found between the current study and prior studies using a crash database approach in analyzing LV-HV interactions. While both approaches found that LV drivers were responsible for the majority of LV-HV interactions, the reasons why these interactions occurred differed with respect to the methodologies used to assess these interactions. For example, the current research found that 22.5% of the LV driver at-fault incidents were cited with the Contributing Factors of Aggressive Driving. In Stuster's (1999) analysis, only 4.3% of the LVs were cited with the driver-related factor "Erratic/Reckless Driving" (it should be noted that Stuster assessed only fatal crashes). Moreover, Hankey et al. (1999) found that 31.1% of the fatal crashes in the FARS database were cited with Aggressive Driving. The results from the current study (22.5%) are within the range reported by Stuster (4.3%) and Hankey (31.1%).

The current research also found that 41.8% of the HV driver at-fault incidents involved a Sideswipe Angle, while 55.1% of the LV driver at-fault incidents involved a Rear End approach. These results differed from Blower's (1998) review of fatal LV-HV crashes. He found that 9.4% of fatal LV-HV interactions, where only the HV driver was cited with a driver related factor, involved a sideswipe angle. Further, Blower's (1998) analysis found that 13.9% of the fatal LV-HV interactions, where only the LV driver was cited with a driver-related factor, involved a rear-end strike. When Council et al. (2003) reviewed all types of LV-HV crashes in North Carolina, they found that 23.2% of the HV driver at-fault crashes involved a sideswipe and 28.5% of the LV driver at-fault crashes involved a rear-end approach. These discrepancies might highlight the differences between analyzing crashes and near crashes and/or the methodologies used analyze the data (i.e., a crash database approach versus a naturalistic or in situ data collection approach).

The results of the current study in conjunction with Hanowski, Keisler, and Wierwille (2004) indicated that LV-HV interactions represent a serious problem. While there were several differences across the three studies, the results consistently showed that LV drivers are more likely to be responsible for the LV-HV interaction than HV drivers. It is believed that the results from the 100-Car, SB, and L/SH studies provide a more complete description of the LV-HV interaction picture. Furthermore, the comparisons among these three studies address the limitations of not having both an instrumented LV and HV.

Scenario classification taken from Thieriez, Radja, & Toth, 2002.