Solving optimal control problems using sigmoid-weighted neural networks

In this study, we are motivated to obtain the solution of optimal controlproblems via universal approximation capability of a single-layer feedfor-ward arti cial neural network. First, we transform the optimal controlproblems into systems of di erential equations. Then, we solve these sys-tems using a single-layer feedforward sigmoid-weighted neural network.Optimal control problemHamiltonian control systemSystem of ordinarydi erential equations Approximate solutions Error back propagation.