back

New release OpenOCL v3.11

We are happy to announce that OpenOCL v3 is ready and released. It contains many improvements with respect to the previous version and also some API changes. If you come from an earlier version see the release notes that contains a list of changes below. Otherwise you can directly go to the download page of OpenOCL v3.11 or pull the latest version from the master branch.

Also note that the URL of the github repository has changed. Update your remote URL with:
git remote set-url origin https://github.com/OpenOCL/OpenOCL.git

Release Notes

The release includes many changes to the API. Some class names, method names, and the way the solver is created are changed. If you come from an earlier version you need to adapt your model/code for this version. Below you find a list of changes. Also look at the examples how they are implemented (and look at the API documentation at openocl.org). This is also the first release to support Octave, and this release will become the first stable release.

API changes:

• Solver.getNLP and Solver.getSolver are replaced by OclSolver(T,system,ocp,options)
• System became OclSystem
• OCP(system) became OclOCP()
• Solver OclSolver takes 1 more argument T, new signature OclSolver OclSolver(T, system, ocp, options)
• There is no explicit time dependents in OCP definition, new signatures: pathCosts(ch, x, z, u, p), arrivalCosts(ch, x, p), pathConstraints(ch, x, p). #83
• In OclSystem: setupEquation became setupEquations #85
• OclSystem and OclOCP: it is possible to implement systems and OCP by in functions, and pass the function handles to the constructors. See Example. This is an alternative way to the existing approach of inheriting from OclOCP and OclSystem. #74
• OclOCP methods pathConstraints, pathCosts have to be declared as Static when inheriting from OclOCP #74
• OclSystem methods setupVariables. setupEquations have to be declared as Static when inheriting from OclSystem #74
• There is no more dependency on z and u in pathConstraints #80
• Slicing for OclVariables works now like in Matlab. #73 In initial guess and solution state/control trajectory are extended in the third dimension, e.g. shape of sol.p is 3x1x(N+1), slice with sol.p(:,1,4:6)

Important algorithm changes:

• Path constraints are enforced for every control interval at the beginning and the end. Before path constraint were only enforced at the end of each control interval. With option path_constraints_at_boundary=false pathconstraint are not enforced on the first state and the final state. This can be used if you have potentially conflicting boundary conditions with your path constraints. #80
• A default control regularization of 1e-6*u.'*u has been introduced. It can be deactivate with option options.disable_regularization = false; #79
• Initial guess is now of type OclValue. The integrator variables for the initial guess are automatically interpolated if no integrator variables has been retreived or set (a warning will be issued in this case). Option options.nlp.auto_interpolation = false disabled auto interpolation.

API new features:

• You can set ocl.setBounds('x', xTraj) where xTraj is a vector with the length of N+1 if x is a state, or N if x is a control variable (with N the number of control interval). #82
• New argument T see OclSolver, can set end time and enables to pass custom discretization points for the OCP. #83
• OclSolver.solve returns times or independent variables as structured variables
• You can set variable limits in OclSystem, #69, VanDerPol
• option opt.system_casadi_mx = true enables the use of casadi MX symbolics for the system

Other changes:

• There is a new example Cartpole Swingup
• Race car example introduce du control
• Unit tests for OclSystem, OclOCP, CollocationIntegrator #72
• Tests are automatically checked by Travis CI
• Subsasgn has been fixed. An issue #77 with octave persists.
• Solving times are reported for profiling in tests #61
• Use of casdiSX symbolics #40
• API docs are in the website #58
• There are warnings issued for compatibility
• Refactorization of the main Variable types
• Better display for variables
• Consistent slicing/indexing of variables
• Solution consists of trajectories for each variable
• Octave support