Can Particle Filter estimate non-linear model transition function parameters

Trying to work with particle filter function and would like to ask if Transition function vdpParticleFilterStateFcn Can use unknown parameters the same way as State Space Model functionality? If it can, can I do the estimation with some ssm analogue?

Answer 1

[Sai Sri Pathuri]

The transition function may use unknown parameters. You can create custom transition function by using source code of vdpParticleFilterStateFcn and add unknown parameters (by defining them as NaN) in the function code.

You may not use the estimate function for estimation because it accepts state space model as input, whereas a particleFilter function creates particle filter object. For estimation, refer correct and predict functions.

[Daniel Margulis]

Thanks! - Will I be able to estimate these unknown parameters in correct/predict inserted functionality? Or I will have to write some complicated likelihood estimation code?

[Sai Sri Pathuri]

I do not think you have to write any complicated code for likelihood estimation. You can use the inbuilt functions such as vdpMeasurementLikelihoodFcn or create a custom function by using the source code for the same.