v3.11.0

LowLevelParticleFilters v3.11.0

Diff since v3.10.1

Merged pull requests:

• add ability to plot covariance in Kalman filter solution plot (#164) (@baggepinnen)

v3.10.1

LowLevelParticleFilters v3.10.1

Diff since v3.10.0

Merged pull requests:

• rm explicit deps (#162) (@baggepinnen)
• more robust detection of inplace measurement for MeasurementOop (#163) (@baggepinnen)

v3.10.0

LowLevelParticleFilters v3.10.0

Diff since v3.9.1

• Performance improvements for large systems
• Support for in-place dynamics and measurement functions in EKF/UKF
• Support for user-defined Jacobians in EKF
• Ability to plot output prediction and prediction error in Kalman-filter solution plot
• New tutorial: Adaptive Neural-Network training

Merged pull requests:

• accept user-defined jacobians in EKF (#158) (@baggepinnen)
• introduce optional non-unit sample interval (#159) (@baggepinnen)
• Performance optimizations for large systems (#160) (@baggepinnen)
• add adaptive neural-network tutorial (#161) (@baggepinnen)

v3.9.1

LowLevelParticleFilters v3.9.1

Diff since v3.9.0

Merged pull requests:

• Add mechanism for rejection of bad sigma points (#155) (@baggepinnen)
• add some tests (#156) (@baggepinnen)
• copy input arrays that are mutated as state (#157) (@baggepinnen)

v3.9.0

LowLevelParticleFilters v3.9.0

Diff since v3.8.0

• Add augmented UKF. This version of the UKF takes explicit noise terms as input to the dynamics and is thus able to handle non-additive noise.
• Move Distributions.jl to a weak dependency and thus reduce the number of transitive dependencies substantially. This should be non-breaking since nothing from Distributions.jl was previously exported. This means that users that only use Kalman filters do not have to depend on Distributions.jl. To handle the most common case of multivariate normal distributions, a LowLevelParticleFilters.SimpleMvNormal type is added.

Merged pull requests:

• CompatHelper: add new compat entry for MonteCarloMeasurements in [weakdeps] at version 1, (keep existing compat) (#151) (@github-actions[bot])
• CompatHelper: add new compat entry for Plots in [weakdeps] at version 1, (keep existing compat) (#152) (@github-actions[bot])
• move Distributions.jl to weak dep (#153) (@baggepinnen)
• add augmented UKF (#154) (@baggepinnen)

Closed issues:

• More flexible noise in UKF (#115)

v3.8.0

LowLevelParticleFilters v3.8.0

Diff since v3.7.1

This version

1. Moves from Requires.jl to package extensions for Plots.jl and MonteCarloMeasurements.jl
2. Removes a significant number of direct and transitive dependencies, corresponding to the following change:

(LowLevelParticleFilters) pkg> rm Requires SciMLBase SeeToDee SimpleNonlinearSolve
    Updating `~/.julia/dev/LowLevelParticleFilters/Project.toml`
  [ae029012] - Requires v1.3.0
  [0bca4576] - SciMLBase v2.59.2
  [1c904df7] - SeeToDee v1.3.0
  [727e6d20] - SimpleNonlinearSolve v2.0.0
    Updating `~/.julia/dev/LowLevelParticleFilters/Manifest.toml`
  [47edcb42] - ADTypes v1.9.0
  [7d9f7c33] - Accessors v0.1.38
  [70df07ce] - BracketingNonlinearSolve v1.1.0
  [7057c7e9] - Cassette v0.3.14
  [38540f10] - CommonSolve v0.2.4
  [a33af91c] - CompositionsBase v0.1.2
  [2569d6c7] - ConcreteStructs v0.2.3
  [187b0558] - ConstructionBase v1.5.8
  [e2d170a0] - DataValueInterfaces v1.0.0
  [a0c0ee7d] - DifferentiationInterface v0.6.22
  [4e289a0a] - EnumX v1.0.4
  [f151be2c] - EnzymeCore v0.8.5
  [e2ba6199] - ExprTools v0.1.10
  [6b7a57c9] - Expronicon v0.8.5
  [9aa1b823] - FastClosures v0.3.2
  [442a2c76] - FastGaussQuadrature v1.0.2
  [6a86dc24] - FiniteDiff v2.26.0
  [f62d2435] - FunctionProperties v0.1.2
  [069b7b12] - FunctionWrappers v1.1.3
  [77dc65aa] - FunctionWrappersWrappers v0.1.3
  [46192b85] - GPUArraysCore v0.2.0
  [3587e190] - InverseFunctions v0.1.17
  [82899510] - IteratorInterfaceExtensions v1.0.0
  [87fe0de2] - LineSearch v0.1.4
  [d8e11817] - MLStyle v0.4.17
  [bb5d69b7] - MaybeInplace v0.1.4
  [be0214bd] - NonlinearSolveBase v1.3.1
  [d236fae5] - PreallocationTools v0.4.24
  [731186ca] - RecursiveArrayTools v3.27.3
  [7e49a35a] - RuntimeGeneratedFunctions v0.5.13
  [0bca4576] - SciMLBase v2.59.2
  [19f34311] - SciMLJacobianOperators v0.1.1
  [c0aeaf25] - SciMLOperators v0.3.12
  [53ae85a6] - SciMLStructures v1.5.0
  [1c904df7] - SeeToDee v1.3.0
  [efcf1570] - Setfield v1.1.1
  [727e6d20] - SimpleNonlinearSolve v2.0.0
  [2efcf032] - SymbolicIndexingInterface v0.3.34
  [3783bdb8] - TableTraits v1.0.1
  [bd369af6] - Tables v1.12.0
  [a759f4b9] - TimerOutputs v0.5.25
  [8ba89e20] - Distributed
  [9fa8497b] - Future

The change should be non breaking, please open an issue if you experience any problems with this change!

Merged pull requests:

• remove unused R2d (#147) (@baggepinnen)
• rm SciML deps (#148) (@baggepinnen)
• rm mcm dep (#149) (@baggepinnen)
• introduce PlotsExt instead of Requires (#150) (@baggepinnen)

v3.7.1

LowLevelParticleFilters v3.7.1

Diff since v3.7.0

Merged pull requests:

• make alpha configurable at each time step (#144) (@baggepinnen)
• JET optimize EKF and switch t from Ref{Int} to Int (#145) (@baggepinnen)
• add tests for differentiability (#146) (@baggepinnen)

v3.7.0

LowLevelParticleFilters v3.7.0

Diff since v3.6.5

• The internals of Kalman filters (all flavors) have been redesigned in order to allow completely allocation free operation when all arrays are static. The performance-tips section of the docs indicate how to set up a filter to achieve this.
• Kalman filters are tested for absence of potential dynamic dispatch using JET.jl. This should hopefully make then suitable for compilation using juliac, and possibly also StaticCompiler.jl
• If you run into method or type errors due to the change of internals, please open an issue.

Merged pull requests:

• add in-place support for dynamics functions in UKF (#139) (@baggepinnen)
• add JET analysis (#141) (@baggepinnen)
• make sure plots are rendered as PNG (#142) (@baggepinnen)
• JET optimize SqKalman (#143) (@baggepinnen)

v3.6.5

LowLevelParticleFilters v3.6.5

Diff since v3.6.4

Merged pull requests:

• test UKF with non-array u (#138) (@baggepinnen)
• CompatHelper: bump compat for SimpleNonlinearSolve to 2, (keep existing compat) (#140) (@github-actions[bot])

Closed issues:

• Does correct!() support missing samples as explained in docs? (#136)

v3.6.4

LowLevelParticleFilters v3.6.4

Diff since v3.6.3

Merged pull requests:

• activate CI cache (#132) (@baggepinnen)
• improve PF loglik estimate (#135) (@baggepinnen)

Closed issues:

• Particle filter with second order Markov model (#133)
• loglik bug for ParticleFilter with resampling (#134)