Graduated Non Convexity

Graduated Non-Convexity (GNC) is an optimization technique used to solve challenging non-convex problems, particularly in areas like pose graph optimization for robotics and computer vision. Current research focuses on improving GNC's efficiency and robustness through adaptive scheduling of the non-convexity parameter and the development of novel algorithms that leverage problem structure, such as decoupling into simpler subproblems or employing linear representations. These advancements lead to faster and more accurate solutions, impacting applications requiring real-time processing of large-scale datasets, ultimately improving the performance of systems like Simultaneous Localization and Mapping (SLAM).