I have the following data

  • Longitudinal acceleration, $a_x^{IMU}$
  • Lateral acceleration, $a_y^{IMU}$
  • Vertical acceleration, $a_z^{IMU}$
  • Yaw angle, $\psi$
  • Yaw rate, $\dot{\psi}$
  • Latitude, $\rightarrow Y$ ( UTM)
  • Longitude, $\rightarrow X$ ( UTM)
  • Speed, $v$

Sampled from an IMU+GPS installed in a car at $10Hz$.

I want to define the motion equations in order to use a Kalman filter.

I've read this post, assuming $a_y=0$ (and $\dot{\psi} = 0)$, I was able to use the filter using these equations $$ \begin{cases} s_{x, t+1} = s_{x,t} + v_{t} \cos(\psi) \, dt + \frac{1}{2} a_{x, t}^{IMU} \cos(\psi) (dt)^2\\ s_{y, t+1} = s_{y,t} + v_{t} \sin(\psi) \, dt + \frac{1}{2} a_{x, t}^{IMU} \sin(\psi) (dt)^2\\ v_{t+1} = v_{t} + a_{x, t}^{IMU} \, dt \\ a_{x, t+1} = a_{x, t}^{IMU} \end{cases} $$ where $(s_x, s_y)$ are observed though $(X, Y)$.

How can I generalize the previous equations with $a_y^{IMU} \ne 0$? Moreover, how can I involve the yaw rate?

From Wikipedia the acceleration of a material point is given by $$ \vec{a} = \frac{dv}{dt}\hat{u}_t + \frac{v^2}{R}\hat{u}_n. $$

Am I right to assume $$ \frac{dv}{dt} = a_x^{IMU} $$ and $$ \frac{v^2}{R} = \dot{\psi} \, v = a_y^{IMU} \ \text{?} $$

In particular, from Performance Vehicle Dynamics the lateral acceleration is given by the sum of two components $$ a_Y = V \dot{\psi} + \dot{V_Y} $$ Why the $a_Y$ above is the sum of two terms? $V \dot{\psi}$ and $\dot{V_Y}$ (contrary to the acceleration of a material point equation)? How this equation is related to the $a_y^{IMU}$ data?

Thanks in advance!


1 Answer 1


Why the $a_Y$ above is the sum of two terms?

From the article:

However, the car also has a horizontal side-slip velocity

The vehicle "lateral" vector $Y$ points perpendicular to the body frame, and thus the direction of travel and accel could be a combination of the forward and lateral velocities. The car is experiencing both "centripetal" accel to make it turn, and may be experiencing lateral accel as it slides perpendicularly.

Ynlike a point mass, which always travels "forward" and accelerates "sideways" and "forwards".

These [two] components are called the tangential acceleration and the normal or radial acceleration

To update when $a_y\neq 0$, you need to update all four of those equations to include a lateral velocity and accel. It will look almost like your accel and vel terms, but will be perpandicular (instead of $cos$ and $-sin$, you have $-cos$ and $sin$ as a unit vector).

And, you'll want to take a close look at $x$ and $y$ in your state equations. It appears from the third / fourth equation that $x$ is body fixed, whereas from the first two $x$ and $y$ are synonymous with Lat/Lon (e.g., earth fixed).

That's fine if $v$ is actually speed, which it is for a forward-accel-only vehicle, but with lateral accel, $v$ needs to be a vector of two velocities. After breaking apart velocity $(v_x(t), v_y(t))$ and accel $(a_x(t), a_y(t))$ and deciding on a common $x$ and $y$ for all equations, it will be easier to rewrite those state transition equations.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.