CONSTRAINED NEAR-OPTIMAL CONTROL USING A NUMERICAL KINETIC SOLVER

摘要

Optimal control problems are challenging to solve even onrnsimple systems. Few problems afford analytical solutionsrnbecause of the boundary valued differential equations. Ki-rnnetic systems, such as robotic linkages, can be complicatedrnto solve because of nonlinearities and degrees-of-freedom.rnA numeric control optimization software, DIDO, is cou-rnpled to a numeric kinetic solver, SimMechanics, withinrnMATLAB. The kinetic model is created directly from arnsolid model assembly eliminating human errors. A pen-rndulum with control saturation is tested to validate satisfac-rntion of theoretical conditions (< 10% optimality residu-rnals, typically < 5%). The numeric method is contrasted torna linear-quadratic-regulator (LQR) and the optimal linearrnstate transfer. A four degree-of-freedom, arm robot pick-rnand-place command is also optimized and realizes a 50%rndecrease in energy used over a ramp to constant velocityrnmaneuver. This coupling obtains near optimal solutionsrnwithout intense, model specific analysis.