Introduction
track和地面打滑的时候,odometry会失效。因此就要估计track和地面的打滑,因此就要建立一个estimating the interaction between a vehicle and the ground的model, 但是通常这些model都有terrain-specific的参数,所以很难应用在unknown environments. 在这个研究中,they propose slip estimation methods targeted toward use in unknown environments
method:
首先定义了four types of slippage based on the slippage direction and maneuver type,也就是建立了一个slip model:
- For straight maneuvers, longitudinal and lateral slippage are derived by approximating the terramechanics slip model.
- For turning maneuvers, longitudinal slippage is derived from an empirical equation for the relationship between slip ratio and input velocity, and lateral slippage is obtained from a regression function
然后就提出了一个slip-compensated odometry, which applies the slip model to the kinematics of a skid-steering vehicle.
总结一下:
In this paper, a simple slip estimation model for tracked vehicles on weak slopes is proposed to be used in combination with the terramechanics- and IMU-based approaches.
We divide slippage into four types, as described below:
-
Longitudinal slippage during straight maneuvers
Longitudinal slippage is estimated based on the terramechanics force interaction model between the track and the terrain. We propose a simplified slippage model that uses approximation to reduce terrain-dependent parameters. -
Lateral slippage during straight maneuvers
Lateral slippage is also estimated based on the terramechanics force interaction model between the track and the terrain. We propose an approximation slippage model with one terrain and robot dependent parameter. -
Longitudinal slippage during turning maneuvers
Skid-steering slippage is a complex phenomenon, and it requires many parameters for estimating the terramechanics based slippage. As mentioned above, the empirical slip estimation formula was confirmed for rigid flat ground. We verify whether the formula can be applied to loose and weak slopes. -
Lateral slippage during turning maneuvers
To estimate lateral slippage while a robot is turning, regression analysis has been used along with training data obtained from the environment. The training data include inertial information and robot position.
Related work
主要有两大类的方法:基于terramechanics和基于IMU的
- 基于terramechanics的方法
This method is mainly used in well-known environments. - 基于IMU的方法
Many researchers employ IMU in combination with sensors-based technologies, such as GNSS, encoders, and cameras.
Kinematics of skid-steering vehicle considering slippage on a plane
V
x
=
v
r
(
1
−
α
r
)
+
v
l
(
1
−
α
l
)
2
V_{x} = \frac{v_{r}(1 - \alpha_{r}) + v_{l}(1 - \alpha_{l})}{2}
Vx=2vr(1−αr)+vl(1−αl)
V
y
=
V
x
t
a
n
β
V_{y} = V_{x}tan \beta
Vy=Vxtanβ
解释一下参数:
V
x
,
V
y
V_{x}, V_{y}
Vx,Vy——the longitudinal and lateral translational velocities in robot coordinate system
这两组参数代表the longitudinal and lateral slippage:
α
r
,
α
l
\alpha_{r}, \alpha_{l}
αr,αl——两条履带的slip ratio;
β
\beta
β——slip angle
v
r
,
v
l
v_{r}, v_{l}
vr,vl——input velocities that can be measured by encoders
在全局坐标系下的kinematics model:
x
˙
=
V
x
c
o
s
θ
−
V
y
s
i
n
θ
\dot{x} = V_{x}cos \theta - V_{y}sin \theta
x˙=Vxcosθ−Vysinθ
y
˙
=
V
x
s
i
n
θ
+
V
y
c
o
s
θ
\dot y = V_{x}sin \theta + V_{y}cos \theta
y˙=Vxsinθ+Vycosθ
θ
˙
=
v
r
(
1
−
α
r
)
−
v
l
(
1
−
α
l
)
2
d
\dot \theta = \frac{v_{r}(1-\alpha_{r}) - v_{l}(1 - \alpha_{l})}{2d}
θ˙=2dvr(1−αr)−vl(1−αl)
下面就需要分straight maneuver和turn maneuver两种情况来分析了
Slippage during straight maneuvering
Force Model of the robot
在直线运动中,有两种力:bulldozing resistance and compaction resistance。本文中只考虑compaction resistance