Global Accelerated Nonconvex Geometric Optimization Methods on SO(3)

• DOI: 10.23919/acc55779.2023.10155919
• Publication Date: 2023-07-03T17:48:03Z
• Authors: Adeel Akhtar; Ricardo G. Sanfelice
• Keywords: Saddle point; Maxima and minima; Hessian matrix; Maxima; Global Optimization; Saddle

This paper proposes global accelerated nonconvex geometric (GANG) optimization algorithms for optimizing a class of functions on the compact Lie group SO(3). Nonconvex is challenging problem because objective function may have multiple critical points, including saddle points. We propose two to escape maxima and points using random perturbations. The first algorithm uses value Hessian perturbations undesired In contrast, second only gradient information efficacy these verified in simulations.