Design of Vehicle Automatic Braking Systems Considering Drivers&#x02019 Braking Characteristics

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Available online at www.ijpe-online.

com

vol. 15, no. 9, September 2019, pp. 2338-2345


DOI: 10.23940/ijpe.19.09.p6.23382345

Design of Vehicle Automatic Braking Systems Considering Drivers’


Braking Characteristics
Zhenhai Gaoa, Tianjun Suna, MuHammad Hassanb,c, and Liupu Wangc,d,e,*
a
College of Automotive Engineering, Jilin University, Changchun, 130022, China
b
Department of Computer Science and Information Technology University of Malakand, KP, Pakista
c
College of Computer Science and Technology, Jilin University, Changchun, 130012, China
d
Cancer Hospital Attached to Xinjiang Medical University, Urumqi, 830000, China
e
Key Laboratory of Symbol Computation and Knowledge Engineering of Ministry of Education, Jilin University, Changchun, 130012, China

Abstract

With the rapid development of advanced driver assistance systems (ADAS), an automatic braking system has become increasingly
important when faced with complicated traffic. The conventional decision-making method for the braking system is focused on security
but lacks consideration of drivers' characteristics, which generates feelings of fear and frustration. We propose an anthropomorphic
braking method based on vehicle kinematic characteristics and dynamic theories to improve the original system and eliminate t he sense of
jerking. Furthermore, on the basis of the traditional time to collision (TTC) algorithm, we consider the random motion of the target
vehicle and develop a dynamic time to collision (DTTC) algorithm to meet different drivers' braking requirements. Therefore, in this
study, the braking model is optimized by a series of driving simulator and real-vehicle braking tests. Finally, the validity and feasibility of
the method are verified through the simulations of different braking modes.

Keywords: automatic braking system; TTC; DTTC; decision-making algorithm

(Submitted on March 10, 2019; Revised on April 12, 2019; Accepted on June 20, 2019)

© 2019 Totem Publisher, Inc. All rights reserved.

1. Introduction

An automatic braking system, as an alternative technology of advanced driver assistance systems (ADAS), can provide an
active braking intervention in cases when the driver faces an emergency risk of collision. Although the traditional braking
algorithm has been widely researched until recently, the model still needs further optimization and improvement.

In early years, Mercedes-Benz found that the main cause of traffic accidents was related to drivers' improper operation.
Accidents caused by human factors accounted for 82% of total incidents, which mainly include 29%, 26%, 19%, and 8% for
rear-end, intersection, lane departure, and changing-lanes collisions, respectively. Only 18% of traffic accidents were caused
by environmental factors or vehicle failure. The database of a German in-depth accident study showed that a complete
braking action of a vehicle when crashing accounted for only 1% of total collision accidents, whereas a partial braking
action when complete braking is needed accounted for approximately 45%. However, accidents resulting from inaction or
minimal braking (with a deceleration of −2m/s2) of a vehicle accounted for more than 50% [1].

In recent years, studies on the mechanism of vehicle longitudinal automatic braking control technology in the world
involve collision time, driver models, safety distance models, or a combination of the three. Minderhoud and Bovy proposed
the concept of time to collision (TTC) [2]. Wachenfeld et al. developed an evaluation index of worst-case collision time
measurement in situation recognition [3]. Chen et al. designed an automatic emergency brake system and other warning
functions on the basis of the early-warning algorithm of "vehicle-to-vehicle" communication by minimizing the calculated
collision prediction time [4]. Jahandideh et al. used an ordered Logit model to establish a model for pedestrians' adventure
behaviour. Results showed that pedestrians prefer to cross the road at the intersection point of the average TTC of

* Corresponding author.
E-mail address: wanglpu@jlu.edu.cn
Design of Vehicle Automatic Braking Systems Considering Drivers’ Braking Characteristics 2339

approximately 6.2s, and 54% of men and 39% of women face TTC risk for less than 3s [5]. Hirose et al. established a model
for the braking behaviour of drivers before rear-end collisions on the basis of the time delay neural network [6]. Gao et al.
established models for warning and braking time prediction on the basis of drivers' habitual braking and braking force
characteristics [7-8]. Yuan et al. established the control strategy of the active collision avoidance system by using the
human-road-vehicle safety distance model. They found that the model can consider the driving safety and road utilization [9].
Yin et al. developed a type of active head restraints in the back-end collision avoidance system in accordance with the
running state of vehicles and the local vehicle braking process of the dynamics and kinematics safety distance calculation
model [10]. These studies have shown that drivers' braking behaviour characteristics are not identical, and the risk
assessment algorithm of standardization and collision avoidance strategy is bound to cause false warnings and unintended
activation and may even interfere with the driver's normal collision avoidance action.

In this study, we attempt to consider different drivers' braking features and then design a humanized braking method for
vehicle automatic braking system on the basis of human reaction time. In this way, we can eliminate the bump sense during
emergency braking by humanizing the braking behaviour in deceleration. Furthermore, the proposed system can improve
the safety of driving and reduce considerable energy loss in traditional braking.

2. Problem Description

An automatic braking system can effectively lower the incidence of accidents, thereby reducing the damage of collision to
drivers. However, the traditional braking algorithm was designed only on the basis of the relative motion of the rigid body
kinematics characteristics and collision theory, and it lacks the understanding and consideration of human braking
behavioural characteristics. Moreover, this situation may lead to a bad driving experience and cause heightened tension
during the period of longitudinal automatic braking control.

The real complex environmental driving condition is composed of three elements: human, vehicles, and traffic
environment [11]. Therefore, all elements that affect vehicle safety whether the collision is avoided or not are integrated.
The mechanical braking distance, drivers' reaction time, and braking system response time play important roles. However,
the braking reaction time of the driver is an important factor that affects the actual braking distance. Moreover, whether the
driver tends to a mild brake or a conventional brake will lead to a different braking distance. Thus, we assume that the
vehicle longitudinal automatic braking control system can function like human beings, such as determining a desired
deceleration or braking pressure before a collision. This system can notify the driver of the potential collision danger, reduce
collision probability, and improve the safety of driving.

The safety judgment of the vehicle driving state mainly depends on the movement of the target vehicle and the driver's
current state. Figure 1 shows the relationship between system structure and key technologies. To realize vehicle longitudinal
automatic braking control, we must solve the following two key technologies and further study the decision-making
algorithm of automatic braking control [12-13].

Figure 1. The system structure and key technologies

1) Trigger time determination. The automatic braking control technology will assume control of the vehicle if the driver
cannot correctly avoid collision in emergencies. Thus, the determination of trigger time becomes increasingly important
when reducing the possibility of vehicle collision. The automatic braking decision-making algorithm can provide an
accurate trigger time for keeping the vehicle far away from the collision danger to avoid or reduce collisions.
2340 Zhenhai Gao, Tianjun Sun, MuHammad Hassan, and Liupu Wang

2) Braking force determination. When the braking deceleration is too small, avoiding collisions is impossible; when the
braking deceleration is too large, it will have a huge impact on the driver and crew of the vehicle. However, in numerous
cases, if the vehicle stops suddenly with a large deceleration, other vehicles on the road are likely to be affected, and this
situation may even lead to other accidents. Thus, the determination of the braking force requires further rigorous calculation.

3. Experimental Analysis

Drivers have different control styles, which results in various braking modes and decelerations. However, data collection
through real-vehicle experiments requires considerable labor and financial resources. Thus, we use a driving simulator with
real drivers in a virtual environment to conduct an experiment. Through this data collection of drivers' different braking
behaviours, we can preliminarily obtain their braking characteristics. Figure 2 shows the experimental results.

(a) Aggressive driver (b) Conventional driver (c) Conservative driver


Figure 2. Results of different drivers with driving simulator experiment

To ensure that the braking behavioural data have enough coverage and discrimination to reflect the vast majority of
drivers' braking characteristics in the process of slowing down, we make a driving simulator test for different types of
drivers. Thus, in accordance with the standard of drivers' classification (aggressive, conventional, and conservative types) in
previous research [14], we select one driver from each type to generate the same condition. Experimental results show that
the ego vehicle can keep an ideal distance in the form of braking based on different driving types and characteristics under
the same condition. Furthermore, in the braking process of different drivers, the aggressive driver has numerous braking
times, and each braking has a high pressure that belongs to a direct deceleration. The conventional driver has minimal
braking times and each braking has a balanced pressure, which belongs to a slow-type deceleration; however, the
conservative driver generally has no braking times. Although the relative distance is decreased when the target vehicle is
slowing down, the ego vehicle can use the motor drag braking or running resistance instead of a mechanical brake due to the
large expectation distance. Finally, we further investigate the correlation between braking modes and speed in the process of
designing an automatic braking decision-making algorithm with a real vehicle.

By analyzing the experimental data from the driving simulator, we find that braking trigger times are different among
various types of drivers. When the vehicle is running at a low speed (approximately 20km/h), the trigger time will be late
and its braking mode is simple (single brake) with minimal deceleration. When the vehicle is running at a high speed
(approximately 80km/h), the trigger time will be early and the braking mode is also simple (single brake) with a large
deceleration. Therefore, the low or high speed of the vehicle is not representative. However, when the vehicle is running at a
moderate speed (approximately 40km/h), the driver adopts a strong form of braking.

4. Automatic Braking Algorithm

An automatic braking system can perform a braking action in accordance with the expected braking characteristics of the
driver based on the vehicle current relative motion state and analysis of the driver's braking intention and mode in the early
stage. The key technical difficulties include the selection of braking trigger time, the calculation of expected deceleration,
and the requirement of braking deceleration.

4.1. Dynamic Time to Collision (DTTC)

The basic requirement of vehicle automatic control is to achieve driving safety. The traditional calculation method for TTC
is shown in Equation (1):

TTC  Drel / Vrel (1)

Where Drel is the relative distance between the vehicles and Vrel represents the relative velocity between the vehicles.
Design of Vehicle Automatic Braking Systems Considering Drivers’ Braking Characteristics 2341

However, this definition of TTC does not consider the speed change of the ego vehicle and target vehicle during acceleration
and deceleration. After reviewing numerous studies, [15-16] we propose the concept of DTTC, which considers the
acceleration or deceleration. Then, we present a further illustration in Figure 3.

Figure 3. Calculation of DTTC by relative motion

As shown in Figure 3, on the basis of the vehicle kinematic relations, the travel distance Dego of the ego vehicle is
calculated using Equation (2):

Dego  Vego  t  aego  t 2 / 2 (2)

Where Vego and aego represent the speed and acceleration of the ego vehicle, respectively. Similarly, the travel distance
Dtar of target vehicle is calculated using Equation (3):

Dtar  Vtar  t  atar  t 2 / 2 (3)

Where Vtar and atar indicate the velocity and acceleration of the target vehicle, respectively. These two vehicles have
their own acceleration in the process of relative motion. Hence, the distance traveled by the ego vehicle is equal to the
distance traveled by the target vehicle plus the current relative distance, as shown in Equation (4):

Dego  Drel  Dtar (4)

Thus, we will obtain the DTTC by combining Equations (3) and (4), as shown in Equation (5):

Vrel  Vrel2  2  arel  Drel


DTTC  (5)
arel

Where arel is the relative acceleration between the vehicles. Four cases are available to obtain the DTTC value:

1) Vrel > 0 and arel > 0;


2) Vrel > 0 and arel < 0;
3) Vrel < 0 and arel > 0;
4) Vrel < 0 and arel < 0.

Therefore, the DTTC value can be calculated using Equation (6):

, Vrel  0 and arel  0



DTTC  Vrel  Vrel2  2  arel  Drel (6)
 , Vrel  0 or arel  0
 arel

4.2. Expected Deceleration

DTTC represents the dynamic time to collision between the vehicles in a moving state. To reflect the drivers' braking
characteristics during vehicle longitudinal automatic control, an ideal model is established as shown in Figure 4.

As shown in Figure 4, the expected deceleration areq represents a required braking deceleration to avoid collision danger.
2342 Zhenhai Gao, Tianjun Sun, MuHammad Hassan, and Liupu Wang

 of the ego vehicle is calculated using Equation (7):


The travel distance Dego

  Vego  t  areq  t 2 / 2
Dego (7)

Figure 4. The braking model for car-following

The calculation of the target vehicle travel distance Dtar  is the same as Equation (3). t represents the time when the
speed of the ego vehicle is reduced to that of the target vehicle. Then, we assume that the ego vehicle achieves the same
speed Vsame as the target vehicle at the end. Thus, t is calculated using Equation (8):

t  (Vego  Vsame ) / areq | Vsame  Vtar | / atar (8)

On the basis of the vehicle following kinematics theory, we can obtain the required deceleration areq using Equations (9)
and (10):

  Dideal  Dtar  Drel


Dego (9)

Vrel  t  (atar  areq )  t 2 / 2  Drel  Dideal  0 (10)

Where Dideal is the expected distance of different drivers. We assume that regardless of whether a driver selects single
or double brake, the ideal distance should be kept to 5 meters due to the different braking modes. Four cases are available to
obtain the areq value:

(1) Vrel > 0 and atar > 0;


(2) Vrel > 0 and atar < 0;
(3) Vrel < 0 and atar > 0;
(4) Vrel < 0 and atar < 0.

Therefore, the areq value can be calculated using Equation (11):

0, Vrel  0 and atar  0


 2
 Vego
, Vrel  0 and atar  0
V 2
 tar
 2( Drel  Dideal )
 atar

areq  Vrel2 (11)
atar  2( D  D ) , Vrel  0 and atar  0
 rel ideal

 Vrel2
2
Vego
atar  or 2 , Vrel  0 and atar  0
 2( Drel  Dideal ) Vtar
  2( Drel  Dideal )
 atar

4.3. A Humanized Braking Method

We can summarize through comprehensive analysis that the braking modes and deceleration are related to the relative
velocity between the vehicles. Therefore, first, regardless of whether the driver decides to brake, the final ideal distance
should be kept to 5 m. Second, we must confirm the relative velocity and braking deceleration. In cooperation with the
Design of Vehicle Automatic Braking Systems Considering Drivers’ Braking Characteristics 2343

technical center of some automobile factories in China, the deceleration determination of the existing automatic braking
system is found to be related to the relative speed. When the relative velocity is greater than 30 km/h, the automatic braking
system will be set to perform a partial brake twice; when the relative velocity is less than 30 km/h, the automatic braking
system will be set to perform a direct brake once. Finally, in designing an automatic braking system to realize a humanized
braking mode, we draw the following conclusions on the basis of the experimental results and theoretical derivations in the
early stage:

(1) Vrel > 30km/h; if areq < −2m/s2, then adec = −2m/s2; if −2m/s2 < areq < −1.5m/s2, then adec = −1.5m/s2;
(2) Vrel < 30km/h; adec = −2m/s2.

Where adec is the braking deceleration. We tend to focus on a humanized braking method during the whole braking
process, but it lacks an emergency situation. In an emergency, the braking deceleration must be set to almost −8m/s2 to meet
the braking requirement and avoid danger. The braking mode will be complex. Figure 5 shows the specific working process
of the automatic braking system.

Figure 5. Working process of the automatic braking system

5. Simulation Tests and Results Analysis

As all driving conditions involved in the automatic braking system during vehicle longitudinal automatic control are
dangerous, a real-vehicle experimental verification is unrealistic to perform directly. However, we use the CARSIM
platform with SIMULINK for joint simulation. Figure 6 shows the process of establishing the simulation model.

Figure 6. Simulation process

5.1. Imitate the Driver with Double Brakes

As previously mentioned, if the relative velocity is greater than 30km/h, we select a double brake. In this way, we will keep
the final distance in the range of 5m. Thus, two simulation tests are conducted.

1) The target vehicle is stationary. In this situation, the target vehicle is stationary at 0km/h, but the ego vehicle keeps
moving at a speed of 40km/h, where v0 is the current speed at time t1 and v1 denotes the current speed at time t2. The first
brake happens between t1 and t2. After t2, the ego vehicle performs a second brake. Figure 7 illustrates the simulation results.

Figure 7. Simulation results when the target vehicle is stationary

2) The target vehicle is moving. In this situation, the target vehicle is moving at v2, and the ego vehicle keeps moving at
2344 Zhenhai Gao, Tianjun Sun, MuHammad Hassan, and Liupu Wang

a speed of 40km/h, where v0 is the current speed at time t1 and v1 represents the current speed at time t2. The first brake
happens between t1 and t2. After t2, the ego vehicle performs a second brake. Figure 8 displays the simulation results.

Figure 8. Simulation results when the target vehicle is moving

5.2. Imitate the Driver with Single Brake

As previously mentioned, if the relative velocity is less than 30km/h, we select a single brake. In this way, we will keep the
final distance in the range of 5m. Thus, two simulation tests are conducted.

1) The target vehicle is stationary. In this situation, the target vehicle is stationary at 0km/h, and the ego vehicle keeps
moving at a speed of 20km/h, where v0 is the current speed of the ego vehicle. The first brake happens at time t1. Figure 9
presents the simulation results.

Figure 9. Simulation results when the target vehicle is stationary

2) The target vehicle is moving. In this situation, the target vehicle is moving at v1, and the ego vehicle keeps moving at
a speed of 25km/h, where v0 is the current speed of the ego vehicle. The first brake happens at time t1. Figure 10 presents the
simulation results.

Figure 10. Simulation results when the target vehicle is moving

6. Real Vehicle Tests and Results Analysis

In this part, we conduct a real vehicle test through Autobox based on the proposed system. Here, we use a flat and straight
road as long as 250m with a moving target vehicle. The results are shown in Figure 11.

(a) Automatic single brake, like an aggressive driver (b) Automatic double brake, like a conventional driver
Figure 11. Results of real vehicle test
Design of Vehicle Automatic Braking Systems Considering Drivers’ Braking Characteristics 2345

As previously mentioned, the deceleration regarded as the dividing line between slight braking and normal braking is
−2m/s2. If the deceleration is less than −2m/s2, it is not considered to be the driver braking characteristics. If the deceleration
is greater than −2m/s2, it is regarded as automatic braking with drivers' braking characteristics. Therefore, when the target
vehicle is slowing down quickly, the host vehicle tends to brake once with a great deceleration of approximately −4m/s2.
When the target vehicle is slowing down gradually, the host vehicle prefers to brake twice with a smooth deceleration of
approximately −2m/s2.

7. Conclusions

This study designed a humanized braking method for vehicle automatic braking systems. First, we used a driving simulator
to obtain different drivers' braking behavioural characteristics. Second, on the basis of the DTTC and the requirement of
expected deceleration, the system was designed under the framework of the vehicle kinematics model and dynamics theory.
Finally, through the simulation tests and real vehicle tests, the validity and feasibility of the method were verified. This
system eliminates the sense of jerking compared with the traditional automatic braking system and also improves comfort
and safety, thereby optimizing the experience of drivers and passengers.

Acknowledgements

This work is supported by the National Science Foundation of China (No. 51775236, 51675224, U1564214), National Key
Research and Development Program (No. 2017YFB0102600), and Natural Science Fund Project of Xinjiang Province (No.
2015211C127).

References
1. X. J. An, M. Wu, and H. He, “A Novel Approach to Provide Lane Departure Warning using Only One Forward-Looking
Camera,” in Proceedings of International Symposium on Collaborative Technologies and Systems, pp. 356-362, IEEE
Computer Society, 2006
2. M. M. Minderhoud, “Extended Time-to-Collision Measures for Road Traffic Safety Assessment,” Accident Analysis and
Prevention, Vol. 33, No. 1, pp. 89-97, 2001
3. W. Wachenfeld, P. Junietz, R. Wenzel, and H. Winner, “The Worst-Time-to-Collision Metric for Situation Identification,” in
Proceedings of Intelligent Vehicles Symposium (IVS), pp. 729-734, 2016
4. Y. R. Chen, J. H. Wan, and J. J. Chen, “Vehicle Collision Prevention Algorithm based on Vehicle Communication,” Telecom
Science, Vol. 32, No. 11, pp. 32-41, 2016
5. Z. Jahandideh, B. Mirbaha, and A. A. Rassafi, “Modelling the Risk Intensity of Crossing Pedestrians in Intersections based on
Selected Critical Time to Collision: A Case Study of Qazvin City,” in Proceedings of Transportation Research Board 96th
Annual Meeting, 2017
6. T. Hirose, M. Gokan, N. Kasuga, and T. Sawada, “Study on Modeling of Driver's Braking Action to Avoid Rear-End Collision
with Time Delay Neural Network,” SAE International Journal of Passenger Cars - Mechanical Systems, Vol. 7, No. 3, pp.
1016-1026, 2014
7. Z. H. Gao, T. Wu, and Z. Zhao, “Vehicle Virtual Following Collision Avoidance Driver Braking Time Model,” Journal of Jilin
University (Engineering Edition), Vol. 44, No. 5, pp. 1233-1239, 2014
8. T. Wu, “A Study on Forward Collision Prevention System Considering the Characteristics of the Driver's Collision Avoidance
Behaviour,” Jilin University, 2014
9. Z. C. Yuan, D. Y. Li, and F. Wu, “DRV Safe Distance Model for Vehicle Longitudinal Active Collision Avoidance,” Journal of
Chongqing University of Technology (Natural Science Edition), Vol. 30, No. 5, pp. 15-19, 2016
10. X. Q. Yin and M. X. Wang, “Research on Safety Distance Mathematical Model of Pro-Active Head Restraint in Rear-End
Collision Avoidance System,” International Journal of Security and Its Applications, Vol. 9, No. 1, pp. 347-356, 2015
11. Y. Lin, C. Wang, J. X. Wang, and Z. Dou, “A Novel Dynamic Spectrum Access Framework based on Reinforcement Learning
for Cognitive Radio Sensor Networks,” Sensors, Vol. 16, No. 10, pp. 1-22, 2016
12. Y. Lin, X. Zhu, Z. Zheng, Z. Dou, and R. Zhou, “The Individual Identification Method of Wireless Device based on
Dimensionality Reduction and Machine Learning,” Journal of Supercomputing, No. 5, pp. 1-18, 2017
13. Y. Lin, C. Wang, C. Ma, Z. Dou, and Z. Ma, “A New Combination Method for Multisensor Conflict Information,” Journal of
Supercomputing, Vol. 72, No. 7, pp. 2874-2890, 2016
14. Z. S. Yu, “Automobile Theory,” Tsinghua University Press, pp. 15-18, 2016
15. S. Takada and H. Kawakami, “Forward Obstacles Collision Warning System based on Deceleration for Collision Avoidance,”
in Proceedings of SICE Annual Conference (SICE), pp. 184-191, 2011
16. Y. L. Chen and K. Y. Shen, “Forward Collision Warning System Considering both Time-to-Collision and Safety Braking
Distance,” Industrial Electronics and Applications, Vol. 6, No. 4, pp. 972-077, 2013

You might also like