Multiple periodic solutions of an equation with state-dependent delay (Q548097)

An object \(x(t)\) moving in \((-w,w)\) controls its position by a signal echoed at \(-w\). Continuing work of H.-O. Walther, the model equation \[ \dot x(t) = g\left[\frac{x(t-1-s(t-1)) + x(t-1)}{2}\right]\tag{1} \] is used, where \(s\) satisfies the implicit equation \[ s(t) = \frac{x(t - s(t)) + x(t) + 2w }{c}, \] \(c\) is the signal speed, and \(g\) is a nonlinearity with negative feedback. It is shown that, for every \(N \in \mathbb{N}\), if \(g\) is sufficiently close to \(-\text{sign}\) (measured by smallness of a parameter \( \varepsilon\)), the equation can have \(N\) (orbitally) different periodic solutions. Formal replacement of \(g\) by \(-\text{sign}\) leads to the equation \[ \dot x(t) = -\text{sign}[x(t-1) + x(t- \gamma)],\tag{2} \] where \( \gamma = 1 + 2w/c\). It is shown that (2) has a corresponding periodic solution \(p=p^{m,n}\) for infinitely many pairs \((m,n) \in \mathbb{N}^2\). Return maps for equations \((1)\) and \((2) \) can be compared, after a process of additional modification defined by periodic extension of the behavior between two neighboring zeroes. It is then shown that for \( p = p^{m,n}\), the local fixed point index of the return map for (2) is nonzero and `survives' the transition to the return map for (1), if \( \varepsilon\) is small enough. The index calculation is based on the description of solutions by their zeroes, resulting in an explicit matrix representation of the derivative at the fixed point. Numerical examples are shown, and the stability of `rapidly oscillating' periodic solutions with, e.g., 6 zeroes per period is conjectured.