Non linear dynamic model parameter identification

3 次查看(过去 30 天)
Hi, i habe a non linear dynamic model of an rov. Since i have the dynamic model equation as well as some parameter of the rov, id like to know if its possible to identify/estimate the unknown parameter of the model by providing measurements ( control Inputs, thruster turns per second, imu data, etc ) with high accuracy? I know there is non linear /greybox systemidentification, but this will not return the unknown params but an estimated System?

采纳的回答

Shubh
Shubh 2024-1-24
Hi Johannes,
Yes, it is possible to estimate the unknown parameters of a non-linear dynamic model for a Remotely Operated Vehicle (ROV) using the available measurements like control inputs, thruster turns per second, and IMU (Inertial Measurement Unit) data. This process is known as parameter estimation or system identification.
In your case, since you already have a dynamic model and some known parameters, you can use a method like non-linear least squares or advanced techniques like the Extended Kalman Filter (EKF) or Particle Filters for parameter estimation. These methods can help you refine the unknown parameters of your model to better match the observed data.
Here's a broad overview of how this can be done:
1. Define the Model: You should have a mathematical representation of your ROV's dynamics. This model should include both known and unknown parameters.
2. Collect Data: Gather measurements from your ROV's operation. This should include inputs to the system (like control signals) and outputs (like position, velocity, which can be derived from IMU data).
3. Choose an Estimation Technique:
  • Non-linear Least Squares: If the model is differentiable and the noise is Gaussian, this method can be effective.
  • Extended Kalman Filter (EKF): Useful for systems with Gaussian noise and where the model can be linearized around the current estimate.
  • Particle Filter: Suitable for systems with non-Gaussian noise or non-linear models that are difficult to linearize.
4. Estimation Process: Use the chosen technique to adjust the unknown parameters in your model so that the model's output closely matches your collected data.
5. Validation: After estimation, validate the model with a separate dataset to ensure the estimated parameters are accurate.
6. Refinement: Based on validation results, you might need to refine your model or estimation technique.
Hope this helps!
  1 个评论
Johannes
Johannes 2024-1-24
Thanks for the detailed answer. I'll give it try. Are the proposed methods somehow existing(toolboxes like system identification) or do i need to build a Kalman filter estimator from scratch for this kind of usecase?

请先登录,再进行评论。

更多回答(0 个)

类别

Help CenterFile Exchange 中查找有关 Linear Model Identification 的更多信息

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by