Computing covariant Lyapunov vectors, Oseledets vectors, and dichotomy projectors: A comparative numerical study (Q1948093)

Two new methods for computing Oseledets vectors are introduced. One method is based on singular value decomposition and the other method is based on dichotomy projectors. The other two existing methods are reviewed and some improvements of them are presented. Finally, a comparative numerical investigation involving all four methods is carried out on three dynamical systems. The first case study is on a dynamical system formed via composition of a sequence of \(8\times 8\)-matrices constructed such that all Oseledets vectors are known at time zero. The second case study is on an eight-dimensional system generated by two hard disks in a quasi-one-dimensional box. The third case study is on a nonlinear model of time dependent fluid flow in a cylinder: the matrices are generated by finite rank approximations of the corresponding time dependent transfer operators. These three cases display the advantages and disadvantages of each of the four considered methods.