Review of the past two weeks

During the last two weeks, I fixed the sparse automatic differentiation support for optimization based solvers in BoundaryValueDiffEq.jl. The previous implementation doesn’t properly utilize the sparsity exploitation in Optimization.jl, wrongly treated the Jacobian prototype as dense matrix and caused dramatic computation overhead. The issue is depicted in issue Allow more autodiff backends when using internal optimization solvers, the related fix is in PR Allow more AD backends in optimization based solvers.

TODOs in next two weeks

In the following two weeks, there are a few TODOs:

Finalize the interface for optimal control problems, make the solving process converging.
Finalize a suitable convention for the cost function in formulating optimal control problems, mainly about the interpolating and integral of control variables.