Centroidal State Estimation

Centroidal state estimation focuses on accurately determining a robot's center of mass position and momentum, crucial for stable and dynamic locomotion. Current research emphasizes developing robust estimation methods, often employing simplified multi-body models and incorporating diverse sensor data, including joint torque measurements and inertial measurement units (IMUs), with algorithms ranging from Extended Kalman Filters to data-driven approaches like Koopman operator theory and reinforcement learning. These advancements improve control precision and enable more agile and adaptable robot behaviors, particularly in challenging scenarios like dynamic legged locomotion and hardware-constrained humanoid robots.