Approximate solution of the target control problem with a nonlinearity depending on one state variable (Q2300699)

The systems are of the form \[ x'(t) = A(t)x(t) + \hat A(t)g(x_1) + B(t)u(t) + f(t) \quad (t_0 \le t \le t_1) \quad x(t_0) = x_0 \tag{1} \] with $x(t) = x(t, x_0, t_0, u)$ and $f(t)$ are $n$-dimensional vector functions. The only nonlinear term $g(x_1) = (g_1(x_1), \dots , g_r(x_1))^T$ depends on $x_1$ only. The vector $x(t) = (x_1(t), \dots , x_n(t))^T$ is the state vector, $u(t) = (u_1(t), \dots , u_m(t))^T$ is the control and $A(t), \hat A(t), B(t)$ are matrix functions. The control constraint is $u(t) \in \mathcal{P}(t)$ where $\mathcal{P}(t)$ is a set valued function whose value at each time is an ellipsoid. The objective is to solve the approximate controllability problem of finding a control $u(t)$ such that \[ x(t_1, x_0, t_0, u) \in \mathcal{X} + B_\mu (0) \] where $B_\mu(0)$ is the ball of radius $\mu$ and center $0,$ $\mathcal{X} = \{x; \phi(x) \le 0\}$ is the target set and the parameter $\mu \ge 0$ is to be minimized; in case $\mu = 0$ the controllability to $\mathcal{X}$ is exact. The authors approximate the nonlinear function $g(x_1)$ by a piecewise linear function and then reduce (1) to a finite number of linear subsystems. They obtain a piecewise quadratic value function and a feedback solution of the problem, and then study approximation to the solution of the original problem (1). This leads to a numerical scheme. The results are applied to two systems: a pendulum attached to one end of a rotating bar and a quadcopter.