Adaptive mesh refinement method for optimal control based on Hermite-Legendre-Gauss-Lobatto direct transcription

Abstract

Direct transcription has been employed to transcribe the optimal control problem into a nonlinear programming problem. This paper presents a trajectory optimization method based on a combination of the direct transcription and mesh refinement algorithm. Hermite-Simpson method has the advantage of reasonable accuracy with highly sparse Hessian matrix and constraint Jacobians, and the pseudospectral method provides spectral accuracy for optimal control problems. The optimal control problem is discretized at a series of Legendre-Gauss-Lobatto points, then the trajectory states are approximated by using local Hermite interpolating polynomials. Thus, the method produces significantly smaller mesh size with a higher accuracy tolerance solution. The derived relative error estimation is then used to trade the number of mesh polynomials degree within each mesh interval with the number of mesh intervals. As a result, the suggested method can produce more small mesh size, requires less computation solution for the same optimal control problem. The simulation experiment results show that the suggested method has many advantages.