This article presents a 2D pose estimation method for an omnidirectional mobile robot with Mecanum wheels, using an extended Kalman filter (EKF) formulated on the Lie group $\mathrm{SO}\left(2\right)$. The purpose is estimate the robot’s position and orientation by fusing angular velocity measurements from the wheel encoders with data from an IMU. Employing Lie algebra, the EKF provides a consistent and compact representation of rotational motion, improving prediction and update steps. The filter was implemented in ROS 1 and validated in simulation using Gazebo, with a reference trajectory and real measurements used for evaluation. The system delivers higher pose estimation precision, validating the effectiveness in rotational maneuvers.