How to Convert State Equations to a State Transition Matrix for a Kalman Filter?

Click For Summary

Discussion Overview

The discussion revolves around converting state equations into a state transition matrix for a Kalman filter, focusing on the mathematical formulation and representation of state dynamics. Participants explore the implications of using differential equations, linearization techniques, and the structure of the state transition matrix in the context of a robotic platform's motion model.

Discussion Character

  • Technical explanation
  • Debate/contested
  • Mathematical reasoning

Main Points Raised

  • One participant presents a state vector and a set of differential equations relating state variables, seeking assistance in forming a state transition matrix.
  • Another participant questions the definition of state progression in terms of differential quantities, expressing confusion over the formulation of the equations.
  • Some participants suggest that the use of cosine and sine functions necessitates linearization around a specific angle, indicating a potential complexity in the equations presented.
  • A participant outlines their system and provides a revised state vector and equations of motion, proposing a state transition matrix but expressing uncertainty about how to incorporate acceleration terms.
  • One participant suggests that creating a Jacobian matrix for the state transition function may be a more appropriate approach.
  • Another participant requests further clarification or details about the entire problem to assist in developing a proper model.

Areas of Agreement / Disagreement

Participants exhibit a lack of consensus on the correct formulation of the state equations and the appropriate method for deriving the state transition matrix. Multiple competing views and approaches are presented, with ongoing uncertainty regarding the treatment of certain terms.

Contextual Notes

Participants express varying assumptions about the definitions and formulations of the state equations, with some suggesting that certain shorthand notations may be misleading. There is also mention of the need for linearization and the potential complexity introduced by nonlinear functions.

Who May Find This Useful

This discussion may be of interest to those working on Kalman filters, robotic motion modeling, and state estimation in dynamic systems, particularly in the context of nonlinear dynamics and state transition matrices.

docsxp
Messages
10
Reaction score
0
I have my state vector containing
$$[X, Y, v_x, v_y, \theta, r, a_x, a_y, b_{\theta}]^T$$

and I have them related by
$$dX = v_x cos \theta - v_y sin \theta\\
dY = v_x sin \theta + v_y cos \theta\\
dv_x = a_x\\
dv_y = a_y\\
d\theta = r\\
dr = 0\\
da_x = 0\\
da_y = 0\\
db_\theta = 0\\
$$

Now I'm actually lost in how to go about in converting them to my state transition matrix representation. Can anyone chime in and help me along please? Thank you.
 
Engineering news on Phys.org
in the Kalman filter, your state progress to the next iteration. how do you define that in terms of your differential quantities ? For instance, your state progressing according to:

\dot{X} = v_x cos(\theta) - v_y sin(theta) ?

I don't quite understand how you have come to the differential quantities.
 
I don't see how you can to do it with the cosine and sine functions there. You'll have to linearize about some theta. Plus, I'm backing @rrdrr8556 in that I also don't understand how you're coming up with those as the differential equations, assuming dX and dY is just bad shorthand for the differential equations of xdot and ydot.

Maybe if you post your entire problem we can help you figure out a proper model for it.
 
timthereaper said:
I don't see how you can to do it with the cosine and sine functions there. You'll have to linearize about some theta. Plus, I'm backing @rrdrr8556 in that I also don't understand how you're coming up with those as the differential equations, assuming dX and dY is just bad shorthand for the differential equations of xdot and ydot.

Maybe if you post your entire problem we can help you figure out a proper model for it.

I am sorry, I should have asked the question in a proper manner. I will outline my system and derivation here.

The system is a robotic platform with a high grade 9-axis (non FOG) IMU and a velocity sensor. The system has 6-dof, but I want to assume 4-dof for simplicity as well as mechanical correction on certain axes.

Hence, my state vector, $$[X, Y, v_x, v_y, a_x, a_y, \omega, \psi, b_\psi]$$ where omega is the turn rate in z-axis (rad/s) and psi is true heading wrt true north. b is the bias (true-gyro). My equations of motion are $$X = X_{0X} + v_x t + \frac{1}{2}a_x t^2$$ $$v_x = v_{0x} + a_x t$$ $$\psi = \psi_0 + \omega_z t$$ and so on for other axes.

Hence my state transition matrix that i came up with for this is $$\left[ \begin{array}{cccc} 1 & 0 & cos\psi & -sin\psi & 0.5(dt)^2 & 0 & 0 & 0 & 0\\
0 & 1 & sin\psi & cos\psi & 0 & 0.5(dt)^2 & 0 & 0 & 0\\
0 & 0 & 1 & 0 & 1 & 0 & 0 & 0 & 0\\
0 & 0 & 0 & 1 & 0 & 1 & 0 & 0 & 0\\
0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 & 0\\
0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0\\
0 & 0 & 0 & 0 & 0 & 0 & 1 & 1 & 0\\
0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0\\
0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1
\end{array} \right]$$

I don't know how to deal with the acceleration term. Any ideas?
 
I think this is not how it should be done, but I should create a Jacobian matrix for F.
 
Can anyone point me in the right direction?
 

Similar threads

  • · Replies 2 ·
Replies
2
Views
2K
Replies
9
Views
7K
  • · Replies 4 ·
Replies
4
Views
2K
  • · Replies 12 ·
Replies
12
Views
2K
  • · Replies 21 ·
Replies
21
Views
3K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 6 ·
Replies
6
Views
1K
  • · Replies 6 ·
Replies
6
Views
8K