A data-driven indirect method for nonlinear optimal control
data-driven approach, indirect method, optimal control, sensitivity analysis
Nonlinear optimal control problems are challenging to solve due to the prevalence of localminima that prevent convergence and/or optimality. This paper describes nearest-neighborsoptimal control (NNOC), a data-driven framework for nonlinear optimal control using indirectmethods. It determines initial guesses for new problems with the help of precomputedsolutions to similar problems, retrieved using k-nearest neighbors. A sensitivity analysistechnique is introduced to linearly approximate the variation of solutions between new andprecomputed problems based on their variation of parameters. Experiments show thatNNOC can obtain the global optimal solution orders of magnitude faster than standardrandom restart methods, and sensitivity analysis can further reduce the solving time almostby half. Examples are shown on optimal control problems in vehicle control and agilesatellite reorientation demonstrating that global optima can be determined with more than99% reliability within time at the order of 10–100 milliseconds.
Tsinghua University Press
Gao Tang,Kris Hauser,A data-driven indirect method for nonlinear optimal control.Astrodyn.2019, 3(4): 345–359.