Available online www.jocpr.

com

Journal of Chemical and Pharmaceutical Research, 2014, 6(3):164-169

ISSN : 0975-7384
Research Article CODEN(USA) : JCPRC5

Application of multiple linear regression model in the performance analysis of

traffic rules
Jingguo Qu*, Yuhuan Cui, Guanchen Zhou and Qingpeng Ding

Qinggong College, Heibei United University, Tangshan, China

_____________________________________________________________________________________________

ABSTRACT

Nowadays, the traffic safety problem has become the focus of attention and making a sound and reasonable traffic
rules and regulations is the trend. The current study mainly discusses the overtaking and lane-changing problem in
highway, the differences between traveling on the left and on the right lane. In order to promote traffic smooth
effectively, overtaking model is constructed. The left lane changing rule considers its influence on human heart, and
then multiple linear regression models. In the country with the cars driving on the left, this model is especially
suitable for the near S shaped highway. By studying the joint influence of the curve radius of curvature, steering
angle and the road friction coefficient on the driving speed of the curve and the obtained speed is within the speed
limit. Therefore, after improving the overtaking model, the left lane vehicle also applies to the lane-changing rule.

Keywords: overtaking model, multiple linear regression model, variance analysis, regression coefficient
_____________________________________________________________________________________________

INTRODUCTION

Nowadays, traffic safety problem has become the focus of attention, and stipulating sound and reasonable traffic
rules is an important part of traffic safety. In most countries, driving on the right side is a common traffic rule to be
abode by, and drivers are required to be on the right lane in highway with many lanes unless they want to overtake
other cars. When overtaking happens, the driver should move to the left lane to surpass other cars and return to the
original lane after it is over. However, it cannot be ignored that some countries require driving on the left side, which
is regarded as a common traffic rule. Therefore, it is of necessity to establish a model to analyze the traffic rules of
driving on the right side, and improve it to make the model play a more effective role in stimulating traffic smooth.
Then make the improved traffic rules applied to the countries with driving on the left side as the common law, and
consider whether it is effective in these countries and optimize the results.

Overtaking model
Model Hypothesis
(1) assume that the vehicle model is of unification, and the vehicles in the team are of the same braking strength and
efficiency, namely the braking distance is only related to its own speed;
(2) ignore the influence of weather, slope, road conditions, and other environmental factors on the speed;
(3) assume that the drivers’ identification of the surrounding environment is consistent.

Constant speed overtaking model

From the perspective of modeling, overtaking can be classified into constant speed overtaking and accelerating
overtaking, and other overtaking behavior is a combination of the two categories. Constant speed overtaking refers
to the overtaking vehicle finishes the overtaking process with constant or near constant speed; accelerating
overtaking refers to the fact that the overtaking vehicle follows the front vehicle, and wait for the suitable moment
to speed up to accomplish the overtaking process. Figure 1 is the sketch map of the overtaking process.

164
Jingguo Qu et al J. Chem. Pharm. Res., 2014, 6(3):164-169
______________________________________________________________________________

Figure 1 Sketch map of overtaking

Model description
Constant speed overtaking is the simplest situation, namely the overtaking car, the surpassed car and the vehicle on
the left lane have the same speed.

Establish a simplified graph shown in Figure 1. The car n  1 is the overtaking car，the car n is the surpassed
car. A-A position marks the place when the overtaking car changes the lane and begins to overtake. B-B position
marks the overtaking car catches up with the surpassed car in the opposition lane. C-C position marks the position
when the overtaking car returns to the original lane.

At the time of t , the car n  1 , n , n - 1 drive at constant speed. Set t C as the total time for the overtaking car to
accomplish the overtaking process, that is， the time required for the car n  1 to move from A-A to C-C .
During this time, the driving distance of the car n  1 is S1  S 2 , the driving distance of the car n is S 4 , the
safe distance between the car n  1 after overtaking and the car n - 1 is S 3 . Then the computation formula of
the overtaking distance S is shown in （1）.
x (t )
S  S1  S 2  S3  n 1  tC  S3 （1）
3.6
In the formula S ——overtaking distance (meter);

t C ——the total time requirement for the car n  1 to accomplish overtaking (second);
xn 1 (t ) ——the speed of the car n  1 (kilometer/hour).

It is known from the previous research that when the time headway is between 2 to 8 seconds，the vehicle will be
influenced and restricted by the front car. At that time, the vehicle has an overtaking requirement, and an overtaking
likelihood may occur. By analyzing the result, the time headway range is given when the overtaking will judgment
begins. When analyzing the vehicle comprehensive movement state and the total feature of the driver, it is
recommended to adopt 3.1 seconds[1].

In the highway, drivers hope to drive at the expected speed. But due to the complexity of the road and traffic
condition, the different vehicle type and performance, together with the drivers’ different temperament, it leads to
the gap of the expected speed between driver and vehicle in the formed people-vehicle unit. Therefore, when the
faster car is driving behind the comparatively slow car, the faster car hopes to main its expected speed, so the
overtaking willingness comes into being.

Overtaking acceleration model

Based on the referred material [2] when the overtaking accomplishes, the safety distance model between the
overtaking car and the car n  1 is:
1
S加  va  (ta  t y )   a  ta2  a  ta  t y
2
S 加 ——the safety distance in the process of accelerating overtaking;

165
Jingguo Qu et al J. Chem. Pharm. Res., 2014, 6(3):164-169
______________________________________________________________________________
v a ——the speed of overtaking car A when it begins to overtake;
t a ——the time for accelerating driving;
t y ——the time for constant driving;
vb ——the constant speed of the surpassed car B.

Based on the pertinent literature and investigation on experienced drivers, only when the spacing distance between
the overtaking car A and the surpassed car Bis bigger than 20m, overtaking is allowed[3].

MULTIPLE LINEAR REGRESSION MODEL

Model preparation
Based on lane changing rules of cellular automata traffic flow model constructed in Problem Two, it applies to the
highway road, no matter the vehicle is on the left or right. The obvious difference is the body's physiological factors.
As is known to all, when we are running, we are used to doing it counterclockwise, this is because the clockwise
running can cause oppression to the heart, and is not conducive to human health. Similarly, when vehicles on the left
lane change the lane, it will produce a part of clockwise driving. Because of inertia, the heart will be on right
avertence. In order to optimize this, we design the approximate "S" shape curve of highway, as shown in figure 2.
We mainly consider the influence of the radius of curvature, steering angle and road friction coefficient on the car
when it drives on the curve. Through the establishment of multiple regression analysis model, we study the common
effect of several factors on the curve driving[4].

Figure 2 highway graph of the near S shaped curve

Table 1 the radius of curvature, steering angle, road friction coefficient and vehicle speed

Steering
speed（ km / h ） Radius of curvature
angle
Friction coefficient
65.45 20 120 0.15
80.04 50 60 0.20
90.42 80 40 0.25
101.02 120 30 0.30
109.64 150 20 0.35
116.89 180 16 0.40
123.01 220 12 0.45
127.28 250 8 0.50

Model Establishment and Solution

Assume dependent variable y and independent variable x1 , x2 , x3 have the following relational expression[5]：
y   0  1 x1   2 x2   3 x3  
Among it, y is observable random variable,  0， 1， 2， 3 are unknown parameter ；  is unobservable
random error, satisfying the condition of E  0, D( )   2 （  2 unknown）, but the random error is  slim，
therefore it can be omitted in actual investigation.

Setting the regression equation by establishing the relationship between vehicle running speed y and radius of
curvature x1 , steering angle x2 , road friction coefficient x3 ：

166
Jingguo Qu et al J. Chem. Pharm. Res., 2014, 6(3):164-169
______________________________________________________________________________

y   0  1 x1   2 x2   3 x3

Evaluate the unknown parameter  0， 1， 2， 3 .

According the least square method, use SPSS software to work out the unknown parameters[6]:

  0  81.546
   0.163
 1

  2  0.188
  3  17.133
Therefore, the equation of linear regression is:
y  81.546  0.163 x1  0.188 x2  17.133 x3

◆ analysis of the model result:

R is
(1)It is known in Table 2, the multi-correlation coefficient of 3 independent variables and dependent variable
2
0.998, the determination coefficient R is 0.995, the adjustedR 2 is 0.99. Due to the adoption of the enter method,
2
there is only one regression model. It is seen in Table 2 that the changing volume of R and determination
2
coefficient R is consistent. At the same time, the three variables can explain the speed dependent variable has
99.5% variation value.
Table 2 Model Summaryb

Std.Error of the
Model R R Square Adjusted R Square
Estimate
.998
1 a .995 .992 1.96778
a. Predictors:(Constant),V4,V3,V2
b. Dependent Variable:V1

(2) Variance analysis [7]

Table 3 is the analysis of variance table of the regression model, F is 283.692, the corresponding Sig.is the actual
significance probability of F . That is the value p . When it is less than 0.05 significance value, the linear relation of
the regression equation is obvious.

Table 3 ANOVAb

Model Sum of Square df Mean Square F Sig.

1 Regression 3295.494 3
1098.498 .000
Residual 15.489 4 283.692 a
3.872
Total 3310.983 7
a. Predictors:(Constant),V4,V3,V2
b. Dependent Variable:V1

(3) Significance testing of regression coefficient

Table 4 Coefficients

Standardized
Unstandardized Coefficients
Model Coefficientts t Sig.
B Std.Error Beta
1 (Constant) 81.546 27.745 2.939 .042
V2 .163 .301 .611 .542 .616
V3 -.188 .045 -.322 -4.174 .014
V4 17.133 205.481 .096 .083 .938
a. Dependent Variable:V1 [8]

From the three variables in table 4: the Sig of radius of curvature, steering angle and road friction coefficient is
respectively: 0.616, 0.014, 0.938.its corresponding Sig. is the actual significance level of the value t , namely the

167
Jingguo Qu et al J. Chem. Pharm. Res., 2014, 6(3):164-169
______________________________________________________________________________
value p , if given   0.05 ，then

p1  0.616   so refuse H 0 , thinking the regression coefficient of independent variable “radius of curvature”
is not obvious.

p2  0.014   so refuse H 0 , thinking the regression coefficient of independent variable “steering angle” is
obvious.

p3  0.938   so refuse H 0 , thinking the regression coefficient of independent variable of “road friction
coefficient” is not obvious.

(4) examine the normal distribution of sample observation:

Table 5 Residuals Statisticsa

Std.Deviatio
Minmum Maximum Mean N
n
Predicted Value
64.7695 129.3852 101.7250 21.69758 8
Residual
-2.10517 2.14368 .00000 1.48750 8
Std.Predicted
-1.703 1.275 .000 1.000 8
Value
-1.070 1.089 .000 .756 8
Std.Residual
a.Dependent Variable:V1

Table 6 relative error

Actual speed Calculated value Relative error%

65.45 64.82 -0.9626
80.09 81.84 2.1850
90.42 91.35 1.0285
101.02 100.61 -0.4059
109.64 108.233 -1.2833
116.89 114.73 -1.8479
123.01 122.86 -0.1219
127.28 129.36 1.6342

From Table 6, the absolute value of the relative error between calculated value and actual value is small; multiple
linear regression analysis meets the following four conditions[9]:
★ linear relationship exists between independent variable and dependent variable;
★observed values are independent from each other;
★residual e obeys normal distribution N (0,  2 ) ;
★ e doesn’t change according to different variable values, that is homogeneity of variance.

Therefore, multiple linear regression model is suitable for the study of curvature radius, steering angle and the
variable relationship between road friction coefficient and vehicle speed.

The model is applied to the country with driving on the left as the rule, such as Britain, its scope of the highway
speed limit is 64.4 / 128.8 km / h [10].The actual value and the calculated values are in the range. Therefore, based
on overtaking model after modification, the left lane vehicles are also suitable for lane changing rules of driving.

In addition, we also consider the driving rules different areas around the world distribution, as is shown in Figure
3.We find that 34% of the countries are driving on the left, 66% of the countries are driving on the right. And
countries with traffic rules of driving on the left are mostly island countries, such as Hong Kong, Australia and other
regions. Therefore, it is related to the geographical position whether it is the rule to drive on the left or right[11].
Through further analysis, most of the countries keep left for the invasion of the British, is only required by Britain
and has not changed. And take Hong Kong and China for example, traffic rules are different, but the difference
between the geographical positions is few. Thus, we conclude that different location of vehicle driving rules do not
have too big effect, therefore, this study does not consider this factor seriously.

168
Jingguo Qu et al J. Chem. Pharm. Res., 2014, 6(3):164-169
______________________________________________________________________________

Figure 3 regional distribution of the traffic rules for different countries in the world

CONCLUSION

To sum up, in the country with the car driving on the left, the improved lane-changing rules meet the function of the
human body comfort, perfect the vehicle lane changing rules based on overtaking model, especially applicable to the
near "S" shaped curve of highway.

future work
Based on the overtaking model established in current study, its main study object is the double-lane changing
situation. Hereby, we aims to promote the overtaking model and apply it to the three lane model.

Assume the three lanes as the carriageway, and vehicles driving in the lanes have the same performance and vehicles
are allowed to overtake other vehicles at any lane. Based on the above condition, establish model by determining the
first lane as the accelerating lane, and the rest two lanes as the driving lane. The stimulated result shows that under
the same condition, three lanes with the managed traffic flow model have similar traffic stream characteristics with
that of the three lane traffic without regulation. However, the system lane-changing frequency and the lane
utilization ratio vary. In this way, it is demonstrated that the three lanes with managed traffic flow model enhance the
drive safety.

Acknowledgements
This research was supported by the National Nature Foundation of China (Grant No. 61170317) and the National
Science Foundation for Hebei Province (Grant No.A2012209043), all support is gratefully acknowledged.

REFERENCES

[1] Shijie Liu. Research of overtaking model on two lane highway in simulation，2007,5.
[2] Hongguo Xu, Biao He.Road traffic accident analysis and reproduction[M]. Police Education
Press,1996,12:14-19.
[3] Wenyin Chen. Regulations of the people's Republic of China road traffic management interpretation and
Application [M],Beijing：People's Communications Press，1996.
[4] Davis L C.Modifications of the optimal velocity traffic model to include delay due to driver reaction
time.www.elsevier.com/locate/physa. 2007, 6
[5] Hongfei Jia. City Road car following behavior neural network simulation model [D]. Changchun: Jilin
University, 2002.
[6] Zhanhong Zhang. Journal of North China Institute of Science and Technology, 2009,6(3):27-30.
[7] You Ren. Driving behavior simulation and emergency driving reliability modeling traffic environment [D ].
Changchun: Jilin University, 2007
[8] Zhiqing Liu,Ru Haige,Biao Gong. Road traffic safety engineering[M]. Beijing: Chemical Industry Press, 2005.
[9]Wenqing Wang,Wuhong Wang,Yonggang Zhong Journal of traffic and Transportation
Engineering,2003,3( 1):72- 75.
[10] Narendra K S, Balakrishnan J. Transportation Research Board of the National Academies[C]. Automatic
Control, IEEE Transactions on.8(6), 2002.
[11] Dianye Zhang. Journal of Southwest Jiao Tong University, 2000, 35(6):319-322.

169