v1.3.1

ModelPredictiveControl v1.3.1

Diff since v1.3.0

• added: show @debug messages in github CI if debug logging is activated
• changed: @debug messages with the getinfo dictionary are never truncated (show_limited=false is no longer necessary)

Merged pull requests:

• added: show @debug message in CI if debug logging is activated (#153) (@franckgaga)

v1.3.0

ModelPredictiveControl v1.3.0

Diff since v1.2.0

• added: show getinfo dictionary in the debug log if activated (instead of only the solution summary of JuMP)
• added: show tips about debug log in the optimization warning and error messages
• debug: fallback if MovingHorizonEstimator arrival covariance update fails (keep the old one)
• changed: covariance matrix inversion with cholesky instead of inv in MovingHorizonEstimator (inv internally uses lu)

Merged pull requests:

• doc: torques constraints AND power constraint in the example (manual) (#150) (@franckgaga)
• debug: fallback if MovingHorizonEstimator arrival covariance update fails (#151) (@franckgaga)
• added: call getinfo when debug logging is activated and not solved (#152) (@franckgaga)

Closed issues:

• Error on execution of doc example (#146)

v1.2.0

ModelPredictiveControl v1.2.0

Diff since v1.1.4

• added: ForwardEuler ODE solver
• added: remove 1 useless .= op in RungeKutta(4)
• added: forcing specialization in the NLP functions by using Vararg{T, N} instead of splatting (see this tip)
• added: skipping computations in extended_predictions when possible
• added: print the solver order for NonLinModel
• changed: bump julia compat to 1.10 (new LTS)
• changed: improve code reuse in NLP functions to improve code maintainability
• test: new tests for ForwardEuler
• doc: info about types that allocate or not

Merged pull requests:

• removing commented out code (#144) (@franckgaga)
• added: new ForwardEuler solver with 0 allocation (#145) (@franckgaga)
• bump julia compat to 1.10 (LTS) (#147) (@franckgaga)
• Added: slightly speed up NonLinMPC and MovingHorizonEstimator (#149) (@franckgaga)

v1.1.4

ModelPredictiveControl v1.1.4

Diff since v1.1.3

• added: linearize! is now allocation free
• added: do not call gc! if nc==0
• added: new mpc.buffer.ΔŨ buffer to reduce the allocations
• changed: preparestate! default d argument as empty vector
• various doc improvements

Merged pull requests:

• changed: preparestate! default d argument as empty vector (#137) (@franckgaga)
• added: do not call gc! if nc==0 (#138) (@franckgaga)
• added: reduce allocations for PredictiveController (#139) (@franckgaga)
• added: linearize! is now allocation free (#141) (@franckgaga)
• Update Project.toml (#143) (@franckgaga)

v1.1.3

ModelPredictiveControl v1.1.3

Diff since v1.1.2

• added: :gc field in getinfo for custom constraints of NonLinMPC
• debug: allow setmodel! on SteadyKalmanFilter/Luenberger when there is no model modification (for changing the weights of a PredictiveController)
• test: validating the :gc fields of getinfo! in the constraint violation tests
• doc: remove Hp parameter in the gc example of the manual (useless)

Merged pull requests:

• added: gc in getinfo (#136) (@franckgaga)

v1.1.2

ModelPredictiveControl v1.1.2

Diff since v1.1.1

• added: save some computations in NonLinMPC with T_lastu field instead of T_lastu0
• changed: moved all controller weights in mpc.weights struct to improve code re-use
• test: improve constraint violation tests (verify the whole horizons, use distinct lower and upper bounds)
• test: new custom constraint violation tests that use Ue argument
• doc: added a custom nonlinear constraint example on the motor power in the manual
• doc: various improvements

Merged pull requests:

• added: save some NonLinMPC computations with T_lastu instead of T_lastu0 for input conversion (#133) (@franckgaga)
• doc: minor correction (#134) (@franckgaga)
• Doc: nonlinear custom constraint example on the motor in the manual (#135) (@franckgaga)

v1.1.1

ModelPredictiveControl v1.1.1

Diff since v1.1.0

• fixed: remove a type-instability in LinMPC and NonLinMPC introduced in v1.1.0
• added: reduce allocations with PredictiveControllers using new fields in mpc.buffer object
• added: improve performance with 4 iszero_X fields for each PredictiveController weight, to skip computations when possible
• changed: moved all controller weights in mpc.weights struct to improve code re-use
• test: new custom constraint violation tests that use Ue argument

Merged pull requests:

• Fix type instability in LinMPC and NonLinMPC (introduced in 1.1.0) (#130) (@franckgaga)
• Changed: moved all controller weights in mpc.weights struct (#131) (@franckgaga)
• Added: iszero_W field in for the 4 weights in ControllerWeights struct (#132) (@franckgaga)

v1.1.0

ModelPredictiveControl v1.1.0

Diff since v1.0.2

• added: support for custom nonlinear inequality constraints gc in NonLinMPC
• added: call JE and gc once in NonLinMPC constructor and show stacktrace if it fails (to guide the user)
• doc: additional details on vector dimensions
• doc: documents the dummy function plot_recipe instead of plot to avoid confusion
• tests: new test for custom nonlinear constraint violation

Merged pull requests:

• Added: support for custom nonlinear constraints gc in NonLinMPC (#118) (@franckgaga)
• Avoid introducing MPC.plot (#120) (@baggepinnen)
• Custom constraints : add tests (#128) (@franckgaga)

Closed issues:

• plot not working (#119)

v1.0.2

ModelPredictiveControl v1.0.2

Diff since v1.0.1

• changed: removed code related to Julia 1.6 compatibility (1.10 is the new LTS)
• doc: add compat info to ModelingToolkit.jl example
• doc: various minor modifications

Merged pull requests:

• Added: details in observability error message, modify CI for new release and MTK exemple debug (#117) (@franckgaga)
• Add compat details on MTK.jl bugfixes (#122) (@franckgaga)
• Doc correction and simple CI fix (#123) (@franckgaga)
• doc: change slmpc speed gain for the real benchmark (#125) (@franckgaga)
• bump (#126) (@franckgaga)

Closed issues:

• Support for dynamic differential programming in a model predictive context (#109)
• MTK example no longer work (#116)
• linearize is throwing an error when using an empty disturbance (#121)

v1.0.1

ModelPredictiveControl v1.0.1

Diff since v1.0.0

• changed: more precise error for LinModel with non-zero D matrix
• doc: MTK example error message for non-strictly proper systems
• doc: MTK example now uses the p argument of NonLinModel
• doc: remove old stuff that no longer applies
• doc: various improvements

Merged pull requests:

• doc: MTK example throw error for non-strictly proper systems (#110) (@franckgaga)
• doc: add the parameter p in f and h function signature (#111) (@franckgaga)
• doc: remove stuff that no longer applies (#113) (@franckgaga)
• bump version (#114) (@franckgaga)

Closed issues:

• Better integration with ModellingToolkit (#106)
• Avoid Matrix Inversion in the MovingHorizonEstimator (#112)