Challenge the future
Delft University of Technology
Efficacy of Traffic Management
Measures: The Influence of Complexity
of Driving Conditions
Dr. R. (Raymond) G. Hoogendoorn, J. (Jaap) Vreeswijk, MSc & Prof. dr. ir. B. (Bart) van Arem and Prof. dr. K. (Karel) A. Brookhuis
2
The influence of complexity on longitudinal driving behavior
Outline
•
Introduction;
•
Introducing a theoretical framework of adaptation effects in
relation to complexity;
•
Method;
•
Results;
•
Conclusion;
3
The influence of complexity on longitudinal driving behavior
Introduction
•
Complexity of driving conditions has
been shown to have a substantial
impact on driving behavior;
•
E.g. Brookhuis et al. (1991), Horrey et
al. (2009);
•
Therefore an impact on the efficacy of
traffic management measures may be
assumed (see for instance
Hoogendoorn et al. (2011);
•
However: how is complexity of driving
conditions actually related to these
adaptation effects in driving behavior?
4
The influence of complexity on longitudinal driving behavior
Introducing a
theoretical
framework
4
•
Governed by an interaction between
driver capability and task demands;
•
Driver capability:
• Driver characteristics;
• Activation level;
• Distraction;
•
Task demands: difficulty of the
driving task;
•
Adverse condition: an imbalance
between task demands and driver
capability occurs;
External circumstances Road design Weather Environ-ment Interactions vehicles Roadside traffic management In-car technology Complexity Static Dyna-mic Driver characteristicsDriver capability Task demands
Mental workload Situational awareness Compensation effects Performance effects Driving behavior
Figure 1: Theoretical framework of adaptation effects in longitudinal driving behavior in relation to complexity
5
The influence of complexity on longitudinal driving behavior
Introducing a
theoretical
framework
2
5
•
To resolve this imbalance:
compensation effects;
•
E.g. Speed reductions, increase in
spacing
•
When insufficient, performance
effects;
•
E.g. perceptual narrowing, longer
inter-decision times, etc;
External circumstances Road design Weather Environ-ment Interactions vehicles Roadside traffic management In-car technology Complexity Static Dyna-mic Driver characteristics
Driver capability Task demands
Mental workload Situational awareness Compensation effects Performance effects Driving behavior
Figure 1: Theoretical framework of adaptation effects in longitudinal driving behavior in relation to complexity
6
The influence of complexity on longitudinal driving behavior
Introducing a
theoretical
framework
3
6Longitudinal
Lateral
microscopic7
The influence of complexity on longitudinal driving behavior
Research questions
7
•
However:
• To what extent does complexity of driving conditions influence compensation effects in longitudinal driving behavior, represented by changes in speed and spacing?
• To what extent does complexity of driving conditions influence perceptual thresholds with regard to relative speed and spacing?
• To what extent does complexity of driving conditions influence the sensitivity of accelerations towards relative speed and spacing?
• To what extent does complexity of driving conditions influence inter-decision times?
8
The influence of complexity on longitudinal driving behavior
Method
8
•
Driving simulator experiment
with a repeated measures
design;
•
Virtual motorway with three
lanes in the same direction;
•
Control condition (normal driving
conditions);
•
Experimental condition (concrete
barriers and narrow lanes);
•
25 participants (mean age:
29.68, SD=6.93, mean driv,
experience: 9.6, SD=7.50);
Figure 2: Driving environment developed for the purpose of the experiment. On the left the control condition is displayed, while on the right the experimental condition is displayed.
3.3
Par ticipants
1
The research population consisted of 25 employees and students of Delft University of Technol-2
ogy (16 male and 9 female participants). The age of he participants varied from 22 to 54 years 3
with a mean age of 29.68 years (SD = 6.93). Driving experience varied from 1 to 35 years with 4
a mean of 9.6 years (SD = 7.50). 5
3.4
Data analysis methods
6
Compensation effects were analyzed through a comparison of the indicators of longitudinal 7
driving behavior (i.e., speed v and spacing s) between the control and the experimental condition 8
using a paired samples t-test with asignificance level of 0.05. 9
In order to determine performance effects represented by changes in the perceptual thresh-10
olds we started with estimating action points in the (Dv, s) plane in a psycho-spacing model 11
using the datafiltering technique described in Hoogendoorn et al. [11]. The basic assumption of 12
the applied method is that a trajectory can be represented by non- equidistant periods in which 13
acceleration is constant. This implies that speed v(t) is a continuous piecewise linear function 14
of time. For instance, let tj for j = 0, ..., M denote the time instants at which the acceleration 15
changes (i.e., the action points). Given these time instants, weaim tofind thepoints yj describing 16
the value of the piecewise linear function at the time instants tj. 17
This provides us with a distribution of action points in the relative speed-spacing (Dv, s) 18
plane. These distributions were compared using a Kolmogorov-Smirnov test with asignificance 19
level of 0.05. Also, in order to be able to compare changes in performance effects in longitudinal 20
driving behavior we estimated the perceptual thresholds through finding the coefficients of the 21
polynomials p(x) in the third degree that fitted the action points p(x(i)) to y(i) in a least squares 22
sense: 23
p(x) = p1x3+ p2x2+ p3x+ p4 (1)
This analysis was performed separately for acceleration reductions and acceleration in-24
creases at the action points. The goodness of fit, which is regarded as an indication for the 25
7 Figure 2: Driving environment developed for the purpose of the experiment. On the left the control condition is displayed, while on the right the experimental condition is displayed.
3.3
Par ticipants
1
The research population consisted of 25 employees and students of Delft University of Technol-2
ogy (16 male and 9 female participants). The age of he participants varied from 22 to 54 years 3
with a mean age of 29.68 years (SD = 6.93). Driving experience varied from 1 to 35 years with 4
a mean of 9.6 years (SD = 7.50). 5
3.4
Data analysis methods
6
Compensation effects were analyzed through a comparison of the indicators of longitudinal 7
driving behavior (i.e., speed v and spacing s) between the control and the experimental condition 8
using a paired samples t-test with asignificance level of 0.05. 9
In order to determine performance effects represented by changes in the perceptual thresh-10
olds we started with estimating action points in the (Dv, s) plane in a psycho-spacing model 11
using the datafiltering technique described in Hoogendoorn et al. [11]. The basic assumption of 12
the applied method is that a trajectory can be represented by non- equidistant periods in which 13
acceleration is constant. This implies that speed v(t) is a continuous piecewise linear function 14
of time. For instance, let tj for j = 0, ..., M denote the time instants at which the acceleration
15
changes (i.e., the action points). Given these timeinstants, weaim tofind thepoints yj describing
16
the value of the piecewise linear function at the time instants tj.
17
This provides us with a distribution of action points in the relative speed-spacing (Dv, s) 18
plane. These distributions were compared using a Kolmogorov-Smirnov test with asignificance 19
level of 0.05. Also, in order to be able to compare changes in performance effects in longitudinal 20
driving behavior we estimated the perceptual thresholds through finding the coefficients of the 21
polynomials p(x) in the third degree that fitted the action points p(x(i)) to y(i) in a least squares 22
sense: 23
p(x) = p1x3+ p2x2+ p3x+ p4 (1)
This analysis was performed separately for acceleration reductions and acceleration in-24
creases at the action points. The goodness of fit, which is regarded as an indication for the 25
9
The influence of complexity on longitudinal driving behavior
Method
2
9
•
Analysis compensation effects in empirical
longitudinal driving behavior through paired
samples t-tests;
•
Analysis of performance effects through:
•
Estimation of action points in relative speed
spacing plane (perceptual thresholds):
Hoogendoorn et al. (2011);
•
Establishing sensitivity of acceleration towards
relative speed / spacing at these action
points:
•
Establishing elapsed time between
10
The influence of complexity on longitudinal driving behavior
Results –Compensation effects
10
•
Substantial and significant effects of complexity on empirical
longitudinal driving behavior;
11
The influence of complexity on longitudinal driving behavior
Results –Performance effects
11
•
Changes in perceptual thresholds;
•
In the complex conditions drivers react predominantly to larger speed
12
The influence of complexity on longitudinal driving behavior
Results –Performance effects
2
12
•
Sensitivity of accelerations to relative speeds and spacing;
13
The influence of complexity on longitudinal driving behavior
Results –Performance effects
3
13
•
In the control condition the
inter-decision times
amounted to 0.58s
(SD=0.46), while in the
experimental condition they
were 0.76s (SD=0.63);
•
Significant difference
between conditions!
14
The influence of complexity on longitudinal driving behavior
Conclusion
14
•
Framework: interaction between
driver capability and task
demands lead to compensation
effects and performance effects;
•
Indeed substantial effects in
empirical longitudinal driving
behavior: compensation effects;
•
Also performance effects due to
complexity:
•
Change in perceptual thresholds;
•
Changes in sensitivity towards
relative speed;
•
Change in inter-decision times;
•
In the evaluation of traffic
management measures these
effects should be taken into
account;
15
The influence of complexity on longitudinal driving behavior 15