v1.13.2

ModelPredictiveControl v1.13.2

Diff since v1.13.1

• changed: new default sparse backend for jacobian and hessian in NonLinMPC and MovingHorizonEstimator

It should be non-breaking in most cases. It's only a performance improvements to potentially reduce the number of colors for differentiation

Merged pull requests:

• Modify MPC setup to use MovingHorizonEstimator (#281) (@franckgaga)
• Fix typo in README for MovingHorizonEstimator (#282) (@franckgaga)
• changed: better jacobian and hessian defaults for the sparse case in NonLinMPC and MovingHorizonEstimator (#283) (@franckgaga)
• bench: add custom constraint tests with hessian=true (#284) (@franckgaga)
• changed: sparse backends default to post-processing (#285) (@franckgaga)

Closed issues:

• Fiddling with the coloring (#280)

v1.13.1

ModelPredictiveControl v1.13.1

Diff since v1.13.0

• debug: prepare AD with rand constant contexts
• debug: only fills non-zeros in ∇²f of @operator

Merged pull requests:

• debug: prepare AD with rand constant contexts (#278) (@franckgaga)
• debug: only fills non-zeros in ∇²f of @operator (#279) (@franckgaga)

Closed issues:

• Support exact Hessians of Lagrangian NonLinMPC and MovingHorizonEstimator (#181)

v1.13.0

ModelPredictiveControl v1.13.0

Diff since v1.12.0

• added: hessian argument in NonLinMPC for exact Hessian of Lagrangian 🎯
• added: hessian argument in MovingHorizonEstimator for exact Hessian of Lagrangian 🎯
• added: trunc_cov function for MHE (to optimize perf. of dot in objective function)
• added: lastu keyword argument to moveinput!

Merged pull requests:

• test: adding trivial tests with oracle=false (#271) (@franckgaga)
• added: lastu keyword argument to moveinput! (#274) (@franckgaga)
• added: hessian keyword argument in NonLinMPC (#275) (@franckgaga)
• added: trunc_cov function for MHE (to optimize perf. of dot in objective function) (#276) (@franckgaga)
• added: hessian keyword argument for MovingHorizonEstimator (#277) (@franckgaga)

v1.12.0

ModelPredictiveControl v1.12.0

Diff since v1.11.1

• added: oracle=true option in NonLinMPC to significantly improve performances under nonlinear constraints 🚀 (default with Ipopt.jl)
• added: oracle=true option in MovingHorizonEstimator to significantly improve performances under nonlinear constraints 🚀 (default with Ipopt.jl)

Merged pull requests:

• Experimentation: testing Ipopt._VectorNonlinearOracle (#229) (@franckgaga)
• doc: minor precision in strictly proper system assumption (#263) (@franckgaga)
• debug: do not warn cond(H̃) in NonLinMPC if Ewt≠0 (#264) (@franckgaga)
• changed: differentiating only non-Inf constraints in NonLinMPC (#267) (@franckgaga)
• added: use Ipopt._VectorNonlinearOracle in MovingHorizonEstimator (#268) (@franckgaga)
• added: migration to MOI.VectorNonlinearOracle (#269) (@franckgaga)
• added: oracle option in NonLinMPC and MovingHorizonEstimator (#270) (@franckgaga)

v1.11.1

ModelPredictiveControl v1.11.1

Diff since v1.11.0

• added: progress keyword argument in sim! to disable progress logs
• added: ill-conditioned QP warning at controller construction
• removed: less useful workload in precompile.jl to accelerate precompilation

Merged pull requests:

• added: progress keyword argument in sim! to disable progress logs (#260) (@franckgaga)
• added: ill-conditioned QP warning at controller construction (#261) (@franckgaga)
• removed: less useful workload in precompile.jl (#262) (@franckgaga)

Closed issues:

• Multi-threaded MultipleShooting and TrapezoidalCollocation (#257)

v1.11.0

ModelPredictiveControl v1.11.0

Diff since v1.10.0

• added: multithreading with MultipleShooting and TrapezoidalCollocation
• test: multithreading with MultipleShooting and TrapezoidalCollocation
• bench: case studies with multithreading on the pendulum

Merged pull requests:

• doc: update pendulum exemple to MTK v10 in the manual (#251) (@franckgaga)
• doc: details on proper systems in LinModel extended help (#252) (@franckgaga)
• bench: only run bench if benchmark appears in label (#254) (@franckgaga)
• added: multithreading with MultipleShooting and TrapezoidalCollocation (#258) (@franckgaga)
• Update transcription.jl (#259) (@franckgaga)

v1.10.0

ModelPredictiveControl v1.10.0

Diff since v1.9.3

• added: covestim in MovingHorizonEstimator now supports SteadyKalmanFilter for a constant arrival covariance $\mathbf{\bar{P}}$

Merged pull requests:

• doc: add arXiv badge (#247) (@franckgaga)
• added: covestim in MovingHorizonEstimator now supports SteadyKalmanFilter (#248) (@franckgaga)
• Bump version (#249) (@franckgaga)

v1.9.3

ModelPredictiveControl v1.9.3

Diff since v1.9.2

• debug: setmodel! now works with LinMPC based on MultipleShooting
• test: new test for setmodel! with MultipleShooting
• doc: using MultipleShooting in the manual on the unstable pendulum system

Merged pull requests:

• doc: using MultipleShooting in the manual for the unstable pendulum system (#245) (@franckgaga)
• debug: setmodel! now works with LinMPC based on MultipleShooting (#246) (@franckgaga)

v1.9.2

ModelPredictiveControl v1.9.2

Diff since v1.9.1

• added: clearer pretty-print of objects with a tree structure
• added: print differentiation backends in NonLinModel, ExtendedKalmanFilter, MovingHorizonEstimator and NonLinMPC
• changed: using $\varepsilon$ instead of $\epsilon$ for MovingHorizonEstimator slack variable (to clearly differentiate from MPC slack)

Merged pull requests:

• added: clearer pretty-print with trees + print differentiation backends (#244) (@franckgaga)

v1.9.1

ModelPredictiveControl v1.9.1

Diff since v1.9.0

• added: piecewise linear manipulated inputs $\mathbf{u}$ with TrapezoidalCollocation(1)
• changed: f̂ now internally handle the operating points
• debug: linearize! on NonLinModel with non-zero operating points now works
• doc: describing the internal of the various prediction and transcription methods

Merged pull requests:

• debug: linearize! on NonLinModel with non-zero operating points now works (#239) (@franckgaga)
• doc: describing the internal of the various prediction and transcription methods (#240) (@franckgaga)
• changed: f̂ now internally handle the operating points (#241) (@franckgaga)
• doc: changing the notation of transcription internals (#242) (@franckgaga)
• added: piecewise linear manipulated inputs with TrapezoidalCollocation(1) (#243) (@franckgaga)