Académique Documents
Professionnel Documents
Culture Documents
2015-01-0591
Published 04/14/2015
Michael Ng
University of Massachusetts
Saied Taheri
Virginia Tech
CITATION: Khanse, K., Pierce, E., Ng, M., and Taheri, S., "Performance Measurement of Vehicle Antilock Braking Systems (ABS),"
SAE Technical Paper 2015-01-0591, 2015, doi:10.4271/2015-01-0591.
Copyright 2015 SAE International
Abstract
Outdoor objective evaluations form an important part of both tire and
vehicle design process since they validate the design parameters
through actual tests and can provide insight into the functional
performances associated with the vehicle. Even with the industry
focused towards developing simulation models, their need cannot be
completely eliminated as they form the basis for approving the
performance predictions of any newly developed model. An objective
test was conducted to measure the ABS performance as part of
validation of a tire simulation design tool. A sample vehicle and a set
of tires were used to perform the tests- on a road with known profile.
These specific vehicle and tire sets were selected due to the
availability of the vehicle parameters, tire parameters and the ABS
control logic. A test matrix was generated based on the validation
requirements. The set of parameters to be measured were extracted
from this test matrix to understand the demand on instrumentation,
data acquisition and post processing of the acquired data. A cost
effective yet novel instrumentation circuit was designed to perform
the measurements. A brief description has been provided on the post
processing performed on the acquired data. The results of this test
have been discussed in details and the variations in results due to
changes in test conditions have been analyzed. These results were
finally used to validate the performance of the developed vehicle
simulation tool.
Introduction
Tractive forces that exist between the tires of a vehicle and the road,
determine to a large extent- the active safety of a vehicle. A number
of systems exist in vehicles today that aim at improving its active
safety. Antilock Braking System (ABS) is one such system. Reports
released by the National Highway Traffic Safety Administration
(NHTSA) show results with statistically significant decrease in
multi-vehicle crashes and fatal pedestrian crashes due to the inclusion
Downloaded from SAE International by Eric Pierce, Wednesday, March 04, 2015
Tire Model
A dynamic tire model was developed that is suitable for the high
frequency events such as ABS and short wavelength disturbances
from the road. The following two sections describe the models
developed to simulate the enveloping property and the tire itself,
within the MATLAB/Simulink environment.
Enveloping Model
If the profile of any given road is considered, it can be seen that it is
never flat, no matter how smooth the road might seem. This
unevenness in the road profile excites the tire in a complex manner.
Due to the built-in flexibility in the geometry of tires, they act as
geometric filters to the unevenness, smoothening out the sharp edges.
As a result, the effective road disturbance acting on the tire is
different from the actual road profile over which the tire rolls. This is
known as the enveloping behavior of a tire. Figure 1, taken from [2],
shows this behavior. The Tandem Elliptical Cam Model as described
by Schmeitz [3] is used to model this enveloping property of the tire.
This model assumes two elliptical cams in tandem, connected by rods
and constrained to move only in the vertical direction, rolling over
the uneven road surface. Figure 2 shows this Tandem Elliptical Cam
Model rolling over a cleat.
ABS Model
The ABS model developed in Simulink consists of two main
components: 1. Control algorithm and 2. Pressure model. Both these
components are equally important to achieve realistic and favorable
results, and hence, a balance needs to be ensured. The ABS unit
requires longitudinal slip and wheel acceleration as inputs from the
tire model. Based on these inputs, the ABS, with the included
pressure model, calculates the braking torque, May, which again
serves as an input to the tire model.
The operating algorithm of an ABS is characterized by the number
and the type of brake states utilized by the ABS to achieve optimal
control of slip and wheel acceleration. In this ABS control logic
designed by Ding et al [5], the primary states for the braking system
include: Initialize, Hold Brake Pressure, Increase Brake Pressure,
Fast Increase Brake Pressure, Step Increase Brake Pressure,
Decrease Brake Pressure, Step Decrease Brake Pressure and Exit
Braking. An ABS system deployed in a vehicle is designed to
control four wheels with a single controller. In a general scenario, it
might be too difficult to calculate control signals on all four wheels
in a single cycle. This is accomplished by the use of controller flags
where one wheel is controlled during a cycle and this is rotated after
each cycle. This way, all four wheels can be controlled at the
desired rate. The selection of brake states are made using a
state-flow based approach where the decisions are made based on a
set of rules and thresholds. Further, the brake pressure which is
regulated by means of electromagnetic valves installed in the
Hydraulic Control Unit (HCU) in the actual system can be regulated
in the simulation through three states: Pressure Increase, Pressure
Hold and Pressure Release. Increase in pressure and decrease in
Downloaded from SAE International by Eric Pierce, Wednesday, March 04, 2015
Vehicle Model
Vehicle Modeling was done using the commercially available
software package CarSim. CarSim is capable of predicting
vehicle performances in response to driver controls (e.g. braking) for
a given environment (e.g. road profile, coefficient of friction). The
models simulate physical tests to allow viewing the results much
more quickly and repeatedly as compared to physical testing. These
models are capable of replicating the system level behavior with high
fidelity. They contain the major effects that determine how the forces
are transferred through the suspension to the chassis. However, they
do not have the details of the linkage connections or structural
compliance. They do not include component details to determine
durability, fatigue or high frequency vibrations. CarSim can accept
Kinematics & Compliances data for a vehicle that can be obtained
from the Kinematics and Compliances (K&C) Rig.
K&C Rig is used to accurately establish the kinematic characteristics
of a vehicle's suspension and steering system geometries, and the
compliance characteristics of the suspension springs, anti-roll bars,
elastomeric bushes and component deformations. It provides the data
required for simulating vehicle performance in terms of ride, impact
isolation, steering and handling. This data combined with the
dimensional and inertia properties of the vehicle is sufficient to
predict the vehicle performances.
Tires
The test tire was selected based on the availability of the
experimental data and available tire parameters for modeling. The
specification of the tire used was 205/60 R15 91V. The tire data was
obtained from a trial version of TNO MF-Tool 6.1 [6] and was
recorded using standard test procedures and experiment measurement
conditions, as described in their documentation. Table 1 enlists some
of the tire parameters that were used as part of modeling.
Table 1. Tire Parameters
Vehicle
A sample vehicle was selected based on the availability of its data for
modeling. The Kinematics and Compliances data of the vehicle was
measured using a K&C Rig. The weight was measured using vehicle
load cells and the center-of-gravity (CG) was calculated using the
same load cells and lift-based measurement procedures. Other vehicle
dimensions were recorded using standard measuring instruments.
Table 2 describes the list of all parameters that were considered for
modeling, from the vehicle used for the test.
Table 2. Vehicle Parameters
Antilock Brakes
The ABS controller used for the model was that from the Antilock
Braking System installed on the sample vehicle. This controller was
previously developed at the same research lab and reference to its
development can be found in [5].
Road
The test was performed on the Smart Road which is a 2.2 mile
controlled-access testing facility at the Virginia Tech Transportation
Institute (VTTI). Broadly, there were two types of road surfaces
present on this road- concrete and asphalt. The concrete road was
seen to be smoother as compared to its rougher asphalt counterpart.
Profiles of these road surfaces were measured using vehicle mounted
road profiling system as part of a previous research project. The
resolution of the scanned profile was taken as 1 mm. A sample
2-dimensional section of this profiled road has been shown in Figure
4, which has a gradient. Further, the test was performed on the
straight section of the Smart Road, since the test involved straight
line braking trials. This straight section of the track is seen to have a
gradient of 2.3. This was seen as a deviation from what has been
considered in the models. In order to negate the effect of this gradient
on test results, it was decided that the test was to be performed uphill
as well as downhill for every test case. An assumption was made that
an average of the test results for braking distances for Uphill and
Downhill trials will be equal to the results obtained for the case if the
road were to be flat.
Downloaded from SAE International by Eric Pierce, Wednesday, March 04, 2015
GPS
A 5V active GPS antenna was used for distance and velocity
measurements. The GPS used was capable of measuring distance at
the rate of 100Hz, with an accuracy of 0.05% and a resolution of
1cm. Likewise, the velocity was also measured at the rate of 100Hz,
with an accuracy of 0.1km/h and a resolution of 0.01km/h. The GPS
antenna which has a magnetic base was mounted on the vehicle roof,
ensuring that atleast 7 satellites were locked during testing at all
times. Figure 5b shows the GPS antenna used.
Figure 4. Road Profile
Test Matrix
Before the test set-up could be designed, a test matrix was developed
to understand the expectations from the validation test. Table 3 shows
the developed test matrix. Using the individual parameters in the test
matrix, the instrumentation set-up was then designed which is
described in the next section.
Table 3. Test Matrix
VBox
The 3rd generation VBox was used for logging the test data from the
brake trigger and the GPS. As stated earlier, the VBox is capable of
recording the data at a rate of 100Hz. The logged data is directly
stored onto a compact flash card to easily transfer to a PC and analyze
using the VBox Tools software. While it simultaneously receives
signals on brake triggering, position as well as velocity, it is easy to
punch the brake trigger event to an exact location, at which point the
velocity is also known. Using this as one reference and the stop
velocity as another, the stopping distance can be calculated from the
position of the vehicle, as all are plotted against time simultaneously.
Figure 5c shows the VBox used.
Part II
Inertial Measurement Unit (IMU)
Test Set-up
Primarily, two instrumentation circuits were designed for measuring
the parameters extracted from the test matrix. The first design, which
included the VBox, was designed from the point of view of
measuring the braking distance and velocity. The other, which
included the Inertial Measurement Unit, was designed from the
point of view of measuring the vehicle accelerations and rotational
rates. They are described in the following two parts:
Part I
Brake Trigger
The brake pedal trigger was used to enable precise measurement of
brake pedal application during the brake test. The trigger events are
recorded at high frequencies of 100Hz. This high frequency provides
a maximum of 10ms gap between the pedal application and when it is
recorded. This time is then automatically included in the calculation
to add the extra distance covered in this short time span. Figure 5a
shows the brake pedal trigger that was used.
Downloaded from SAE International by Eric Pierce, Wednesday, March 04, 2015
Testing
To evaluate the straight line antilock braking of the vehicle, the initial
velocity of the vehicle was considered to be 65km/h (40.3mi/h).
Starting at this velocity, the vehicle was brought to a complete stop
using ABS, simulating panic braking situations. As has been
mentioned earlier, the gradient of the road was taken into
consideration by performing the runs uphill and downhill, and then
averaging the two test runs. Every run was performed 3 times to
check for repeatability in the results. The vehicle was maintained at
the test weight throughout, with considerations of driver, co-driver
and gas in the vehicle. This was confirmed after the test when the
vehicle was weighed again. Standard testing protocols were followed
while testing. Quick checks were made after each run if the data was
being recorded appropriately. A test log was maintained to make
critical notes such as what test was performed on which run- along
with a time stamp, so as to make data processing easier. With two
different circuits recording data simultaneously, it was made sure that
data miss-match does not occur. Syncing the times of these two
clocks precisely with each other ensured this. It was made sure that
an experienced driver was performing the test. The most critical
aspects of driving were maintaining the initial speed up to the
required mark and to simulate the panic braking condition with a
completely depressed brake pedal. A post-test analysis showed that
the initial test speed error was maintained at lower than 1.1% of the
originally required speed for all test runs. This confirmed that the
variations due to human factors were kept to a minimum, so as not to
affect the test results in any manner. The tests were performed in
favorable environmental conditions. Wind speeds were ensured to be
lower than 10-15mph so as not to have a considerable effect on the
vehicle to affect its braking performance. Apart from the above
mentioned details, all standard safety procedures were followed while
performing these tests.
Downloaded from SAE International by Eric Pierce, Wednesday, March 04, 2015
Braking Distance
Figure 8 shows a comparison of the braking distances for the
validation results vs simulation results:
1.
2.
3.
4.
5.
1.
2.
The plot shows the simulation result traces the test result
closely. A correlation coefficient of 0.9923 is observed between
the two plots. This is considered as a very highly correlated
data.
Velocity
Downloaded from SAE International by Eric Pierce, Wednesday, March 04, 2015
Acceleration
6.
7.
1.
References
1.
2.
It is seen that the simulation result generally has the same trend
as the validation test result.
2.
3.
3.
4.
4.
5.
Ding N., Wang W., Yu G., Zhang W., Xu X., Nenggen D.,
Weida W., Guizhen Y., and Wei Z., Research and Validation
of the adaptive control strategy for ABS based on experimental
knowledge, automotive engineering, vol 31, no. 1, 2009.
6.
7.
8.
Conclusions
1.
2.
3.
4.
Downloaded from SAE International by Eric Pierce, Wednesday, March 04, 2015
The Engineering Meetings Board has approved this paper for publication. It has successfully completed SAEs peer review process under the supervision of the session organizer. The process
requires a minimum of three (3) reviews by industry experts.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or
otherwise, without the prior written permission of SAE International.
Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE International. The author is solely responsible for the content of the paper.
ISSN 0148-7191
http://papers.sae.org/2015-01-0591