Gradient Estimate for Continuous-Time Linear Dynamical Systems

In this paper we compare two different methodologies for computing the gradient of a linear, continuous-time dynamical system with no parametric uncertainties nor unknown inputs. Such a gradient is computed by performing the partial derivative of a performance function w.r.t. the input of the system. The first of the methodologies uses the model of the dynamical system to construct the gradient; whereas the second one avails of the differentiation of the performance function and the system’s input. We compare both approaches via a numerical simulation.