Revisiting Implicit Differentiation for Learning Problems in Optimal Control

In contrast, we directly evaluate the matrix equations which arise from applying variable elimination on the Lagrange multiplier terms in the (differential) KKT system.