Fast escape in incompressible vector fields (Q1650731): Difference between revisions

Suppose you find yourself blindfolded in a the middle of the ocean and want to escape the unit disk centered at your starting point \(0\). Simply swimming straight is impossible due to the stream of the ocean, while swimming along the stream is not the best strategy. The author presents a mathematical description of this problem. Let \(\text{v}: \mathbb{R}^2 \to \mathbb{R}^2\) be a smooth, divergence-free vector field (stream of the ocean) satisfying the condition \(c_1 < \|\text{v}\| < c_2\) for some \(0<c_1<c_2<\infty\) and \(\text{v}^{\perp}\) be the orthogonal vector field. Given a curve \(\gamma: \mathbb{R}_+ \to \mathbb{R}^2\), denote by \(l\) the length to escape the unit disk centered at \(0\), that is, \[ l = \inf_{t>0}\, \biggl\{ \int_0^t \|\dot \gamma (s) \| \, ds \;: \;\|\gamma(t)-\gamma(0)\| = 1 \biggr\}, \] where \(\|\cdot\|\) means the standard Euclidean norm in \(\mathbb{R}^2\). The length \(l\) of a flow curve \(\dot \gamma(t) = \text{v}(\gamma(t))\) can be as long as \(l \sim c_2/c_1\), the same is true for the orthogonal flow curve. The author shows that a combination of \(\text{v}\) and \(\text{v}^{\perp}\) does strictly better: there always exists a curve flowing first along \(\text{v}^{\perp}\) and then along \(\text{v}\) which escapes the unit disk before reaching the length \(l \sim \sqrt{4\pi c_2/c_1}\). Moreover, if the escape length of \(\text{v}\) is uniformly \(\sim c_2/c_1\), then the escape length of \(\text{v}^{\perp}\) is uniformly \(\sim 1\) (allowing for a fast escape from the current). Finally, an elementary quantitative Poincaré-Bendixson theorem is established. In conclusion, the reviewer takes the liberty to suggest the following generalization of the problem. In fact, we deal with the control system \[ \dot \gamma(t) = u_1\text{v}(\gamma(t)) + u_2\text{v}^{\perp}(\gamma(t)), \eqno(1) \] where \(u_i=u_i(t)\) are scalar inputs. The author uses controls of a special type: \(u_1 \equiv 0\), \(u_2 \equiv 1\) switches to \(u_1 \equiv 1\), \(u_2 \equiv 0\) once. He remarks that the case of \(k \geq 1\) switches between \(\text{v}\) and \(\text{v}^{\perp}\) is an open problem. It seems natural to consider the extremal problem \(l \to \min\) for system (1) with general inputs \(u_i(t)\) satisfying, for instance, the condition \(u_1^2+u_2^2 \leq 1\).