589665
HFSXXX10.1177/0018720815589665Human FactorsModeling Pedal Application Types
Modeling Types of Pedal Applications Using a Driving Simulator Yuqing Wu, Linda Ng Boyle, University of Washington, Seattle, Daniel McGehee, Cheryl A. Roe, University of Iowa, Iowa City, Kazutoshi Ebe, and James Foley, Toyota Collaborative Safety Research Center, Ann Arbor, Michigan Objective: The aim of this study was to examine variations in drivers’ foot behavior and identify factors associated with pedal misapplications. Background: Few studies have focused on the foot behavior while in the vehicle and the mishaps that a driver can encounter during a potentially hazardous situation. Method: A driving simulation study was used to understand how drivers move their right foot toward the pedals. The study included data from 43 drivers as they responded to a series of rapid traffic signal phase changes. Pedal application types were classified as (a) direct hit, (b) hesitated, (c) corrected trajectory, and (d) pedal errors (incorrect trajectories, misses, slips, or pressed both pedals). A mixed-effects multinomial logit model was used to predict the likelihood of one of these pedal applications, and linear mixed models with repeated measures were used to examine the response time and pedal duration given the various experimental conditions (stimuli color and location). Results: Younger drivers had higher probabilities of direct hits when compared to other age groups. Participants tended to have more pedal errors when responding to a red signal or when the signal appeared to be closer. Traffic signal phases and locations were associated with pedal response time and duration. The response time and pedal duration affected the likelihood of being in one of the four pedal application types. Conclusion and Application: Findings from this study suggest that age-related and situational factors may play a role in pedal errors, and the stimuli locations could affect the type of pedal application. Keywords: pedal misapplications, foot trajectory, driver behavior, pedal application types
Address correspondence to Linda Ng Boyle, Department of Industrial and Systems Engineering, University of Washington, Box 352650, Seattle, WA 98195, USA; e-mail: [email protected]. HUMAN FACTORS Vol. XX, No. X, Month XXXX, pp. 1–13 DOI: 10.1177/0018720815589665 Copyright © 2015, Human Factors and Ergonomics Society.
Introduction The pedal positions in a vehicle have evolved substantially over the years to minimize the likelihood of a pedal misapplication. The brake pedal is typically placed higher than the accelerator pedal (Pollard & Sussman, 1989), and the centers of the brake and gas pedals are separated by roughly one foot length (Pollard & Sussman, 1989; Snyder, 1976). The introduction of brake-to-shift interlocks (BTSI) in the early 1980s minimized pedal misapplication events from occurring at the beginning of the drive cycle (Pollard & Sussman, 1989), specifically for vehicles with automatic transmissions. With BTSI, the brake pedal must always be depressed while the driver shifts into any gear. This requirement prevents any abrupt shifting of the vehicle (Amagasa, 1991) and has been shown to reduce about 83% of unintended acceleration events (Schmidt, 1993). However, Schmidt, Young, and Ayres (1999) later showed that the system could not prevent pedal errors while the car was in motion. Despite the decades of research in pedal designs, drivers still misapply the pedal. In 2004, approximately 6,000 to 7,000 crashes recorded in police reports in Japan were associated with pedal errors (Institute for Traffic Accident Research and Data Analysis, 2004). Tran, Doshi, and Trivedi (2011) referred to pedal errors as the situation when the driver mistakenly presses the wrong pedal or does not press any pedal at all. Rogers and Wierwille (1988) showed that about 0.2% of participants’ foot movements in a simulator study resulted in the wrong pedal or both pedals being pressed, and Tomerlin and Vernoy (1990) showed that 1 out of 169 drivers would continue to step on the wrong pedal. In general, most pedal applications (braking or accelerating) are successful in that no safety situations occur. However, some drivers do
Downloaded from hfs.sagepub.com at Freie Universitaet Berlin on June 24, 2015
2 Month XXXX - Human Factors
confuse the pedals or may slip as they transition from one pedal to another. Such pedal misapplications are rare, but a safety concern still exists as to why drivers may confuse or misapply the pedals (Pollard & Sussman, 1989; Schmidt, 1989). Past studies have focused on driver’s pedal response time (Lee, McGehee, Brown, & Reyes, 2002; Muttart, 2005), but few researchers have examined the trajectory of the foot as it transitions from the accelerator to the brake pedal. Police reports have been used to identify factors associated with crashes that involve pedal misapplication, including driver’s age, gender, and environmental context (Schmidt et al., 1999; Schmidt, Young, Ayres, and Wong, 1997). However, these data sets do not provide information on changes in foot movements. They also rely heavily on self-reports with little information available prior to the incident. More recently, data from integrated cameras, sensors, simulators, and instrumented vehicles provide insights on specific driver errors (Cantin, Blouin, Simoneau, & Teasdale, 2004; Kimura & Shinohara, 2012; Tran et al., 2011). For example, in a study on drivers’ pedal applications using colored stimuli with a stop signal that would interrupt participants’ responses, Kimura and Shinohara (2012) showed that older drivers had greater variability in foot movements when compared to young adults. However, the study did not specifically show a relationship between variability and wrong pedal placement. Foot movements have also been examined using a vision-based approach to examine drivers’ intent to brake (McCall & Trivedi, 2007; Tran et al., 2011). These studies provide an understanding of foot behaviors from different perspectives. Because the strength and duration of the pedal press may also affect foot movements (Schmidt, 1989), additional studies could provide insights on drivers’ pedal touch movement for different events as well as insights on variability in drivers’ pedal responses. There are also different classifications of pedal applications. Rogers and Wierwille (1988) defined three types of pedal errors—serious, catch, and scuff—and they also suggested four potential sequences of these applications. Tran et al. (2011), Schmidt and Young (2010), and
Young, Heckman, and Kim (2011) considered two types: slip/miss and wrong pedal. Past studies provide definitions of pedal errors, but a complete examination of foot movements should also include types of correct pedal placements as well. In an actual traffic environment, drivers need to respond to the color of the traffic signal (red, green, or amber), which becomes routine with more skill-based processing. That said, the driver might face situations when the signal transitions to another color at an unanticipated time. The proximity of the traffic signal can affect the quickness of the drivers’ pedal decisions, and any hesitation may result in severe consequences (Doshi, Tran, Wilder, Mozer, & Trivedi, 2012). There could also be significant differences in the mounting of traffic lights given various road types; traffic lights could be mounted on poles situated on street corners, hung over each lane of the road, or installed in the middle of an intersection. Thus, it is possible that different pedal applications may occur depending on whether the traffic lights are to the right, in front of, or to the left of the driver. That is, the external environmental could potentially influence drivers’ pedal applications. The goal of this study is to identify characteristics of foot behavior and patterns associated with pedal misapplication using a real vehicle placed in a simulated environment (or a hybrid-driving simulator). The two main research questions addressed in this study include the following: •• Do differences exist in foot behaviors given traffic signal color and location? •• If differences exist, what factors are associated with a larger number of pedal errors?
Method
This study was designed to purposely overexpose drivers to pedal misapplication situations. Given that pedal misapplications are rare events, repeated scenarios were used to identify the likelihood of these situations. It was surmised that the ability to switch between responses is an executive function that has the closest relationship to pedal misapplications. Tasks that require switching of responses might impair human performance (Gilbert & Shallice, 2002; Philipp &
Downloaded from hfs.sagepub.com at Freie Universitaet Berlin on June 24, 2015
Modeling Pedal Application Types
3
Figure 1. Camera view and placement of reference line.
Koch, 2005), and depending on the most recent events, the likelihood of a pedal misapplication could increase (Doshi et al., 2012). Participants
Healthy adults with a valid driver’s license in the state of Iowa (United States) were invited through e-mail and web announcement to attend a 2-hr study. There were 62 drivers who participated in this study. Of these, 19 were removed because either the participant did not pass the screening procedure or portions of the experiment data were missing due to machine malfunctions. Hence, data for 43 subjects (20 males and 23 females) were used in the forthcoming analyses. The subjects ranged in age from 18 to 83 years old. A 2012 Toyota Camry was used in this current study, with none of the participants reporting owning this model year as their primary vehicle. Two participants did report owning a Toyota Camry of model year 2000 and 2011, respectively. Each participant was given a practice drive prior to the study and was compensated $40 for his or her time. Apparatus
The study used an instrumented 2012 Toyota Camry XLE secured in a stationary position as the driving simulator. The foot pedals were decoupled from control of the vehicle, and sensors were installed to record pedal presses and touches made by the participant. A 70-in. screen was mounted over the hood of the car, on which the driving scenario was projected. Three foot-well cameras were installed to acquire video of foot movements on the pedals. In addition, cameras were installed
to acquire videos of the forward view of the driving scenario and the view over the driver’s shoulder. Another video camera was positioned on the driver’s face. Since the 2012 Toyota Camry was the only vehicle used for the study, the pedal was always in the same position (see Figure 1 for the pedal configuration). The pedals were 6.80 cm apart horizontally from edge to edge. The vertical separation for the brake and gas pedal was 5.80 cm. The gas pedal was 10.5 cm long and 4.0 cm wide. The brake pedal was 6.4 cm long and 11.50 cm wide. Road Scenario
The scenarios used (Figure 2a) were based on a previous study conducted in Japan (Abe et al., 2012) and included traffic around the participant’s vehicle, which moved at a moderate level of service. Horizontal traffic signals included red, amber, and green phases (Figure 2b), and the traffic signal could be located in the left, middle, or right of the center lane. The signals appeared at varying lengths of time ranging from 0.76 to 1.13 s (median = 1.04 s). The time interval between the two signals ranged from 0.33 to 2.30 s (median = 0.53 s). The drive encompassed 78 green traffic phases, nine red phases, and seven amber phases (for a total of 94 phases). Participants were not told how many green, red, or amber phases were included in the study. The focus of this study was on the driver’s immediate pedal applications to the red and green phases of the stimuli. The task in this study was similar to Kimura and Shinohara (2012), who conducted a simple
Downloaded from hfs.sagepub.com at Freie Universitaet Berlin on June 24, 2015
4 Month XXXX - Human Factors
Figure 2. Example of (a) the simulated road environment and (b) the signal light states.
pedal response task using colored stimuli. The road condition (three lanes in one direction) was similar to Doshi et al. (2012), but their traffic signals would appear in the same location. In this current study, the signal varied across the three lanes and may appear closer or farther from the participant. That is, the signal appeared to be closer if the signal was larger in size on the projection screen and farther away if the signal was smaller. The modification was used to address our research question related to differences in foot behavior given traffic signal location. Only one road type was used because Ising, Droll, Kroeker, D’Addario, and Goulet (2012) showed that road condition was not a significant factor related to the delay in driver’s response. The simulated environment was set up to be in “adaptive cruise control” mode, and the vehicle “steered” itself in the middle lane of the highway. In other words, the scene would move toward the driver at a predetermined rate. This scenario was used given the findings of Abe et al. (2012) that showed no differences in steering angle and lane deviations among drivers in a study on pedal operation. Procedure
Participants signed an informed consent form upon arriving at the facility and agreeing to participate. They then filled out a questionnaire that included demographic and driving information. Instructions for the main study were provided along with a 2-min practice drive to familiarize participants with the simulator and the driving scenarios. In the main study, drivers were asked to respond as rapidly and accurately as possible in response to several traffic signal phases
(stimuli). Their pedal selection was based on the following signal states: •• If the signal was green, they were to press the accelerator pedal. •• If the signal was amber, they were to do nothing. •• If the signal was red, they were to apply the brake pedal.
According to Lococo, Staplin, Martell, and Sifrit (2012), most errors occur when the driver’s foot transitions from the floor toward the pedal(s). Hence, during the “null” state (in between signal lights), the driver was instructed to return his or her feet to a position behind a reference line located on the floor pan (Figure 1). There was no communication between the experimenter and the participant during the study. Data Analysis Independent Variables
The experiment included two between-subject factors: age group (four levels) and gender (two levels), and three within-subject factors: signal color (two levels: green and red), signal lane position (three levels: left, middle, and right), and signal distance (two levels: farther and closer). Drivers’ age and gender are included in the analysis given prior associations with pedal error incidents (Lococo et al., 2012). Participants were recruited from four age groups: young (
HFSXXX10.1177/0018720815589665Human FactorsModeling Pedal Application Types
Modeling Types of Pedal Applications Using a Driving Simulator Yuqing Wu, Linda Ng Boyle, University of Washington, Seattle, Daniel McGehee, Cheryl A. Roe, University of Iowa, Iowa City, Kazutoshi Ebe, and James Foley, Toyota Collaborative Safety Research Center, Ann Arbor, Michigan Objective: The aim of this study was to examine variations in drivers’ foot behavior and identify factors associated with pedal misapplications. Background: Few studies have focused on the foot behavior while in the vehicle and the mishaps that a driver can encounter during a potentially hazardous situation. Method: A driving simulation study was used to understand how drivers move their right foot toward the pedals. The study included data from 43 drivers as they responded to a series of rapid traffic signal phase changes. Pedal application types were classified as (a) direct hit, (b) hesitated, (c) corrected trajectory, and (d) pedal errors (incorrect trajectories, misses, slips, or pressed both pedals). A mixed-effects multinomial logit model was used to predict the likelihood of one of these pedal applications, and linear mixed models with repeated measures were used to examine the response time and pedal duration given the various experimental conditions (stimuli color and location). Results: Younger drivers had higher probabilities of direct hits when compared to other age groups. Participants tended to have more pedal errors when responding to a red signal or when the signal appeared to be closer. Traffic signal phases and locations were associated with pedal response time and duration. The response time and pedal duration affected the likelihood of being in one of the four pedal application types. Conclusion and Application: Findings from this study suggest that age-related and situational factors may play a role in pedal errors, and the stimuli locations could affect the type of pedal application. Keywords: pedal misapplications, foot trajectory, driver behavior, pedal application types
Address correspondence to Linda Ng Boyle, Department of Industrial and Systems Engineering, University of Washington, Box 352650, Seattle, WA 98195, USA; e-mail: [email protected]. HUMAN FACTORS Vol. XX, No. X, Month XXXX, pp. 1–13 DOI: 10.1177/0018720815589665 Copyright © 2015, Human Factors and Ergonomics Society.
Introduction The pedal positions in a vehicle have evolved substantially over the years to minimize the likelihood of a pedal misapplication. The brake pedal is typically placed higher than the accelerator pedal (Pollard & Sussman, 1989), and the centers of the brake and gas pedals are separated by roughly one foot length (Pollard & Sussman, 1989; Snyder, 1976). The introduction of brake-to-shift interlocks (BTSI) in the early 1980s minimized pedal misapplication events from occurring at the beginning of the drive cycle (Pollard & Sussman, 1989), specifically for vehicles with automatic transmissions. With BTSI, the brake pedal must always be depressed while the driver shifts into any gear. This requirement prevents any abrupt shifting of the vehicle (Amagasa, 1991) and has been shown to reduce about 83% of unintended acceleration events (Schmidt, 1993). However, Schmidt, Young, and Ayres (1999) later showed that the system could not prevent pedal errors while the car was in motion. Despite the decades of research in pedal designs, drivers still misapply the pedal. In 2004, approximately 6,000 to 7,000 crashes recorded in police reports in Japan were associated with pedal errors (Institute for Traffic Accident Research and Data Analysis, 2004). Tran, Doshi, and Trivedi (2011) referred to pedal errors as the situation when the driver mistakenly presses the wrong pedal or does not press any pedal at all. Rogers and Wierwille (1988) showed that about 0.2% of participants’ foot movements in a simulator study resulted in the wrong pedal or both pedals being pressed, and Tomerlin and Vernoy (1990) showed that 1 out of 169 drivers would continue to step on the wrong pedal. In general, most pedal applications (braking or accelerating) are successful in that no safety situations occur. However, some drivers do
Downloaded from hfs.sagepub.com at Freie Universitaet Berlin on June 24, 2015
2 Month XXXX - Human Factors
confuse the pedals or may slip as they transition from one pedal to another. Such pedal misapplications are rare, but a safety concern still exists as to why drivers may confuse or misapply the pedals (Pollard & Sussman, 1989; Schmidt, 1989). Past studies have focused on driver’s pedal response time (Lee, McGehee, Brown, & Reyes, 2002; Muttart, 2005), but few researchers have examined the trajectory of the foot as it transitions from the accelerator to the brake pedal. Police reports have been used to identify factors associated with crashes that involve pedal misapplication, including driver’s age, gender, and environmental context (Schmidt et al., 1999; Schmidt, Young, Ayres, and Wong, 1997). However, these data sets do not provide information on changes in foot movements. They also rely heavily on self-reports with little information available prior to the incident. More recently, data from integrated cameras, sensors, simulators, and instrumented vehicles provide insights on specific driver errors (Cantin, Blouin, Simoneau, & Teasdale, 2004; Kimura & Shinohara, 2012; Tran et al., 2011). For example, in a study on drivers’ pedal applications using colored stimuli with a stop signal that would interrupt participants’ responses, Kimura and Shinohara (2012) showed that older drivers had greater variability in foot movements when compared to young adults. However, the study did not specifically show a relationship between variability and wrong pedal placement. Foot movements have also been examined using a vision-based approach to examine drivers’ intent to brake (McCall & Trivedi, 2007; Tran et al., 2011). These studies provide an understanding of foot behaviors from different perspectives. Because the strength and duration of the pedal press may also affect foot movements (Schmidt, 1989), additional studies could provide insights on drivers’ pedal touch movement for different events as well as insights on variability in drivers’ pedal responses. There are also different classifications of pedal applications. Rogers and Wierwille (1988) defined three types of pedal errors—serious, catch, and scuff—and they also suggested four potential sequences of these applications. Tran et al. (2011), Schmidt and Young (2010), and
Young, Heckman, and Kim (2011) considered two types: slip/miss and wrong pedal. Past studies provide definitions of pedal errors, but a complete examination of foot movements should also include types of correct pedal placements as well. In an actual traffic environment, drivers need to respond to the color of the traffic signal (red, green, or amber), which becomes routine with more skill-based processing. That said, the driver might face situations when the signal transitions to another color at an unanticipated time. The proximity of the traffic signal can affect the quickness of the drivers’ pedal decisions, and any hesitation may result in severe consequences (Doshi, Tran, Wilder, Mozer, & Trivedi, 2012). There could also be significant differences in the mounting of traffic lights given various road types; traffic lights could be mounted on poles situated on street corners, hung over each lane of the road, or installed in the middle of an intersection. Thus, it is possible that different pedal applications may occur depending on whether the traffic lights are to the right, in front of, or to the left of the driver. That is, the external environmental could potentially influence drivers’ pedal applications. The goal of this study is to identify characteristics of foot behavior and patterns associated with pedal misapplication using a real vehicle placed in a simulated environment (or a hybrid-driving simulator). The two main research questions addressed in this study include the following: •• Do differences exist in foot behaviors given traffic signal color and location? •• If differences exist, what factors are associated with a larger number of pedal errors?
Method
This study was designed to purposely overexpose drivers to pedal misapplication situations. Given that pedal misapplications are rare events, repeated scenarios were used to identify the likelihood of these situations. It was surmised that the ability to switch between responses is an executive function that has the closest relationship to pedal misapplications. Tasks that require switching of responses might impair human performance (Gilbert & Shallice, 2002; Philipp &
Downloaded from hfs.sagepub.com at Freie Universitaet Berlin on June 24, 2015
Modeling Pedal Application Types
3
Figure 1. Camera view and placement of reference line.
Koch, 2005), and depending on the most recent events, the likelihood of a pedal misapplication could increase (Doshi et al., 2012). Participants
Healthy adults with a valid driver’s license in the state of Iowa (United States) were invited through e-mail and web announcement to attend a 2-hr study. There were 62 drivers who participated in this study. Of these, 19 were removed because either the participant did not pass the screening procedure or portions of the experiment data were missing due to machine malfunctions. Hence, data for 43 subjects (20 males and 23 females) were used in the forthcoming analyses. The subjects ranged in age from 18 to 83 years old. A 2012 Toyota Camry was used in this current study, with none of the participants reporting owning this model year as their primary vehicle. Two participants did report owning a Toyota Camry of model year 2000 and 2011, respectively. Each participant was given a practice drive prior to the study and was compensated $40 for his or her time. Apparatus
The study used an instrumented 2012 Toyota Camry XLE secured in a stationary position as the driving simulator. The foot pedals were decoupled from control of the vehicle, and sensors were installed to record pedal presses and touches made by the participant. A 70-in. screen was mounted over the hood of the car, on which the driving scenario was projected. Three foot-well cameras were installed to acquire video of foot movements on the pedals. In addition, cameras were installed
to acquire videos of the forward view of the driving scenario and the view over the driver’s shoulder. Another video camera was positioned on the driver’s face. Since the 2012 Toyota Camry was the only vehicle used for the study, the pedal was always in the same position (see Figure 1 for the pedal configuration). The pedals were 6.80 cm apart horizontally from edge to edge. The vertical separation for the brake and gas pedal was 5.80 cm. The gas pedal was 10.5 cm long and 4.0 cm wide. The brake pedal was 6.4 cm long and 11.50 cm wide. Road Scenario
The scenarios used (Figure 2a) were based on a previous study conducted in Japan (Abe et al., 2012) and included traffic around the participant’s vehicle, which moved at a moderate level of service. Horizontal traffic signals included red, amber, and green phases (Figure 2b), and the traffic signal could be located in the left, middle, or right of the center lane. The signals appeared at varying lengths of time ranging from 0.76 to 1.13 s (median = 1.04 s). The time interval between the two signals ranged from 0.33 to 2.30 s (median = 0.53 s). The drive encompassed 78 green traffic phases, nine red phases, and seven amber phases (for a total of 94 phases). Participants were not told how many green, red, or amber phases were included in the study. The focus of this study was on the driver’s immediate pedal applications to the red and green phases of the stimuli. The task in this study was similar to Kimura and Shinohara (2012), who conducted a simple
Downloaded from hfs.sagepub.com at Freie Universitaet Berlin on June 24, 2015
4 Month XXXX - Human Factors
Figure 2. Example of (a) the simulated road environment and (b) the signal light states.
pedal response task using colored stimuli. The road condition (three lanes in one direction) was similar to Doshi et al. (2012), but their traffic signals would appear in the same location. In this current study, the signal varied across the three lanes and may appear closer or farther from the participant. That is, the signal appeared to be closer if the signal was larger in size on the projection screen and farther away if the signal was smaller. The modification was used to address our research question related to differences in foot behavior given traffic signal location. Only one road type was used because Ising, Droll, Kroeker, D’Addario, and Goulet (2012) showed that road condition was not a significant factor related to the delay in driver’s response. The simulated environment was set up to be in “adaptive cruise control” mode, and the vehicle “steered” itself in the middle lane of the highway. In other words, the scene would move toward the driver at a predetermined rate. This scenario was used given the findings of Abe et al. (2012) that showed no differences in steering angle and lane deviations among drivers in a study on pedal operation. Procedure
Participants signed an informed consent form upon arriving at the facility and agreeing to participate. They then filled out a questionnaire that included demographic and driving information. Instructions for the main study were provided along with a 2-min practice drive to familiarize participants with the simulator and the driving scenarios. In the main study, drivers were asked to respond as rapidly and accurately as possible in response to several traffic signal phases
(stimuli). Their pedal selection was based on the following signal states: •• If the signal was green, they were to press the accelerator pedal. •• If the signal was amber, they were to do nothing. •• If the signal was red, they were to apply the brake pedal.
According to Lococo, Staplin, Martell, and Sifrit (2012), most errors occur when the driver’s foot transitions from the floor toward the pedal(s). Hence, during the “null” state (in between signal lights), the driver was instructed to return his or her feet to a position behind a reference line located on the floor pan (Figure 1). There was no communication between the experimenter and the participant during the study. Data Analysis Independent Variables
The experiment included two between-subject factors: age group (four levels) and gender (two levels), and three within-subject factors: signal color (two levels: green and red), signal lane position (three levels: left, middle, and right), and signal distance (two levels: farther and closer). Drivers’ age and gender are included in the analysis given prior associations with pedal error incidents (Lococo et al., 2012). Participants were recruited from four age groups: young (
