28-11-2012, 04:17 PM
Vehicle State Estimation for Roll Control System
[attachment=41532]
Abstract
Methods of estimating key vehicle states for roll
stability control systems, including roll rate, roll angle, and
lateral velocity, are investigated in this paper. Roll angle and
roll rate estimators based on the two different vehicle models,
3-DOF (degree-of-freedom) and 1-DOF models, are evaluated
and compared in both linear and non-linear regions using
CarSim, a non-linear simulation software for vehicle dynamics
control and integration. The effects of using roll rate sensor
measurement in estimation are discussed. This paper also
presents a new scheme of lateral velocity estimation, which
compensates for errors in lateral acceleration and longitudinal
speed measurements.
INTRODUCTION
VEHICLE roll stability has become an important issue
in recent years for vehicles with relatively high center
of gravity (CG), such as sport utility vehicles (SUVs) and
trucks [1]-[3]. According to a recent NHTSA study (2004),
non-collision rollover accidents account for approximately
11% of vehicle fatalities in USA even though non-collision
rollover accidents constitute only 2.3% of all vehicle
accidents [4],[5]. In order to reduce the number of these fatal
accidents, impending vehicle rollover can be warned to
drivers or can be mitigated by activating adequate chassis
control systems with knowledge of vehicle roll stability
conditions [6],[7]. Vehicle rollover mitigation systems can
show satisfactory performance when the roll stability
conditions are accurately known to the systems. To identify
vehicle roll stability conditions precisely, it is advantageous
to know the vehicle’s roll rate and roll angle since they are
the most important states in vehicle roll dynamics.
1-DOF Vehicle Model
In addition to the 3-DOF vehicle model, the following 1-
DOF vehicle model is used for estimating vehicle roll angle
and roll rate [5]. This model considers only vehicle roll
motion as opposed to the 3-DOF model, which takes into
account vehicle lateral and yaw motions as well as roll
motion. However, the 1-DOF model provides more robust
results in non-linear maneuvers since it is less sensitive to
non-linear tire dynamics than the 3-DOF model. In [5], the
roll angle is first estimated by integration of roll rate
information from a gyro and then is estimated by a static
function of measured lateral acceleration. The final roll
angle estimation is a combination of these two based on the
1-DOF model.
CONCLUSION
This paper presents methods of estimating vehicle states
for roll stability enhancement and warning systems. The
estimator based on the 1-DOF vehicle model is more robust
than the estimator based on the 3-DOF vehicle model in the
non-linear region because the 1-DOF model is less sensitive
to non-linear dynamics. In addition, the accuracy of roll
angle and roll rate estimation is significantly degraded
without roll rate measurement in the non-linear region. It is
also shown that the lateral velocity estimation, which is
based on the integration of inertial sensor measurements, can
be improved by including the effect of roll motion and better
estimation of vehicle speed.
[attachment=41532]
Abstract
Methods of estimating key vehicle states for roll
stability control systems, including roll rate, roll angle, and
lateral velocity, are investigated in this paper. Roll angle and
roll rate estimators based on the two different vehicle models,
3-DOF (degree-of-freedom) and 1-DOF models, are evaluated
and compared in both linear and non-linear regions using
CarSim, a non-linear simulation software for vehicle dynamics
control and integration. The effects of using roll rate sensor
measurement in estimation are discussed. This paper also
presents a new scheme of lateral velocity estimation, which
compensates for errors in lateral acceleration and longitudinal
speed measurements.
INTRODUCTION
VEHICLE roll stability has become an important issue
in recent years for vehicles with relatively high center
of gravity (CG), such as sport utility vehicles (SUVs) and
trucks [1]-[3]. According to a recent NHTSA study (2004),
non-collision rollover accidents account for approximately
11% of vehicle fatalities in USA even though non-collision
rollover accidents constitute only 2.3% of all vehicle
accidents [4],[5]. In order to reduce the number of these fatal
accidents, impending vehicle rollover can be warned to
drivers or can be mitigated by activating adequate chassis
control systems with knowledge of vehicle roll stability
conditions [6],[7]. Vehicle rollover mitigation systems can
show satisfactory performance when the roll stability
conditions are accurately known to the systems. To identify
vehicle roll stability conditions precisely, it is advantageous
to know the vehicle’s roll rate and roll angle since they are
the most important states in vehicle roll dynamics.
1-DOF Vehicle Model
In addition to the 3-DOF vehicle model, the following 1-
DOF vehicle model is used for estimating vehicle roll angle
and roll rate [5]. This model considers only vehicle roll
motion as opposed to the 3-DOF model, which takes into
account vehicle lateral and yaw motions as well as roll
motion. However, the 1-DOF model provides more robust
results in non-linear maneuvers since it is less sensitive to
non-linear tire dynamics than the 3-DOF model. In [5], the
roll angle is first estimated by integration of roll rate
information from a gyro and then is estimated by a static
function of measured lateral acceleration. The final roll
angle estimation is a combination of these two based on the
1-DOF model.
CONCLUSION
This paper presents methods of estimating vehicle states
for roll stability enhancement and warning systems. The
estimator based on the 1-DOF vehicle model is more robust
than the estimator based on the 3-DOF vehicle model in the
non-linear region because the 1-DOF model is less sensitive
to non-linear dynamics. In addition, the accuracy of roll
angle and roll rate estimation is significantly degraded
without roll rate measurement in the non-linear region. It is
also shown that the lateral velocity estimation, which is
based on the integration of inertial sensor measurements, can
be improved by including the effect of roll motion and better
estimation of vehicle speed.