Magnetometer-less State Estimation for Mobile Robots using Cascaded Kalman Filters

State estimation is crucial in robotics. Traditional outdoor localization methods typically rely on GPS and inertial measurement units (IMUs). However, GPS is ineffective indoors and magnetometers in IMUs are unreliable in miniature robots due to magnetic interference. This paper proposes a magnetometer-less state estimation method using a cascaded extended Kalman filter (EKF). The approach involves two EKFs: the first based on the unicycle model and the second for IMU bias estimation. Zero-velocity update methods (ZUPT) and gradient descent optimization further enhance accuracy. The proposed algorithm was validated through simulation and experiments, demonstrating improved state estimation with minimal sensor drift.