An efficient adjoint computational method based on lifted IRK integrator and exact penalty function for optimal control problems involving continuous inequality constraints

摘要

<p style='text-indent:20px;'>Adjoint methods applied to solve optimal control problems (OCPs) have a restriction that the number of constraints shall be less than that of optimization variables. Otherwise, they are less efficient than the forward methods. This paper proposes an efficient adjoint method to solve OCPs for index-<inline-formula><tex-math id="M1">\begin{document}$ 1 $\end{document}</tex-math></inline-formula> differential algebraic systems with continuous-time inequality constraints. The continuous-time inequality constraints are not discretized on time grid but transformed into integrals and penalized in the cost through an exact penalty function. Thus, all the constraints except for box constraints on optimization variables can be removed. Furthermore, a lifted implicit Runge-Kutta (IRK) integrator with adjoint sensitivity propagation is employed to accelerate the function and gradient evaluation procedure. Based on a sensitivity update technique, the number of Newton iterations involved in forward simulation can be reduced to one. Besides this, Lagrange interpolation is applied to approximate the states not on collocation points such that integrals in the penalty function can be evaluated on the same grid for forward simulation. Complexity analysis shows that, for the proposed algorithm, computation involved in the sensitivity propagation is comparable to that of forward one. Numerical simulations on the optimal maneuvering a Delta robot demonstrate that the computational speed of the proposed adjoint algorithm is comparable to that of our previous one, which is based on the lifted IRK integrator and forward sensitivity propagation.