remove the unused argument df

[baggepinnen]

df for APF is only used in smoothing, where it is used in the call

 logpdf(df, xb[m,t+1], f(xf[n,t],u[t],p,t), t)

to the custom method

@inline Distributions.logpdf(d::Distribution,x,xp,t) = logpdf(d,x-xp)

This PR thus has to

1. Not remove the argument from the docs completely
2. Require that a custom distribution or a custom logpdf function is supplied instead
3. Update the beetle tutorial to show how this is done

[Copilot]

Pull Request Overview

This PR removes the unused argument “df” from AdvancedParticleFilter and updates the code and documentation accordingly. Key changes include:

• Removing the “df” argument from test and documentation examples.
• Introducing the MissingDistribution type and updating constructors to default to it.
• Adding an error check in the smoothing function to enforce that a proper noise density is provided for smoothing.

Reviewed Changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
test/runtests.jl	Removed the unused “df” argument from AdvancedParticleFilter calls.
src/utils.jl	Introduced the MissingDistribution type as a placeholder for a missing noise density.
src/smoothing.jl	Added an error check to ensure that smoothing fails with a clear message if MissingDistribution is used.
src/PFtypes.jl	Updated the constructor to remove the dynamics_density argument and set its default to MissingDistribution.
docs/src/index.md	Updated documentation example to remove the “df” argument.
docs/src/beetle_example.md	Modified plotting parameters for consistency in the beetle tutorial example.

Comments suppressed due to low confidence (2)

test/runtests.jl:296

• There is a typo in the comment ('Simuate' should be 'simulate').

x,u,y = LowLevelParticleFilters.simulate(apf,T,du) # Simuate trajectory using the model in the filter

src/PFtypes.jl:178

• Update the constructor documentation to clarify that a valid noise density must be supplied for smoothing since the default MissingDistribution() will trigger a runtime error in that context.

AdvancedParticleFilter(N::Integer, dynamics::Function, measurement::Function, measurement_likelihood, initial_density; p = SciMLBase.NullParameters(), threads = false, kwargs...) = AdvancedParticleFilter(N, dynamics, measurement, measurement_likelihood, MissingDistribution(), initial_density; kwargs...)

[Copilot]

[nitpick] Consider using a more specific custom exception instead of a generic error to allow downstream code to catch and handle this condition in a more controlled manner.

Suggested change

	df isa MissingDistribution && error("In order to perform smoothing, you must provide a noise density for the dynamics. Use the constructor `AdvancedParticleFilter(N, dynamics, measurement, measurement_likelihood, dynamics_density, initial_density; kwargs...)` to do this")
	df isa MissingDistribution && throw(MissingDynamicsDensityError("In order to perform smoothing, you must provide a noise density for the dynamics. Use the constructor `AdvancedParticleFilter(N, dynamics, measurement, measurement_likelihood, dynamics_density, initial_density; kwargs...)` to do this"))