Adjoint Method

A classical technique for solving optimal control problems of constrained dynamical systems. By introducing Lagrange multipliers, it efficiently computes gradients of the objective with respect to control variables, with memory cost growing linearly in the number of iteration steps.