Initial condition issues on iterative learning control for non-linear systems with time delay (Q4780593)

Asymptotic properties of a conventional learning algorithm are examined for a class of nonlinear systems with time delay in the presence of initial function errors: \[ \left\{\begin{aligned} \dot x_k(t) &=f(x_k(t),x_k(t-\tau),t)+ B(x_k(t),x_k(t-\sigma),t)u_k(t),\\ y_k(t)&=g(x_k(t),t), \end{aligned}\right. \] where \(t\) is the time in the operation interval \([0,T]\) and \(k\) is the number of operation cycles. For \(t\in[0, T]\) and for all \(k, x_k(t)\in\mathbb R^n, u_k(t)\in\mathbb R^r\) and \(y_k(t)\in\mathbb R^m\) are the state, control input and output of the system, respectively. Both \(\tau > 0\) and \(\sigma > 0\) are constant time delays. For \(t\in [-\mu,0]\), \(\mu = \max\{\tau,\sigma\}\), \(x_k(t) = \psi_k(t)\) and \(\psi_k(t)\) is the initial function of the system. Given a desired trajectory \(y_d(t)\), \(t\in [0, T]\), the objective is to find a control input such that the system output follows the desired trajectory. A conventional learning algorithm takes the form of \[ u_{k+1}=u_k(t)+L(y_k(t), t)(\dot y_d(t)- \dot y_k(t)), \] where the learning gain \(L(\cdot,\cdot)\) is piecewise continuous and bounded on \(\mathbb R^m\times [0, T]\). It is shown that the tracking performance can be poor due to an initial function shifting when a conventional learning algorithm is applied. An initial rectifying action is introduced for performance improvement and a proof is provided for analysing its robustness and convergence against initial function errors. The initial rectifying action is finite and used over a specified interval. It is shown to be effective in the improvement of tracking performance, in particular robustness and uniform convergence. The results are further extended to systems with multiple time delays. An example is given and computer simulations are presented to demonstrate the performance of the proposed approach.