System and method for recovering inertial sensor data using asynchronous prediction buffer on a microcontroller.

Abstract

The invention proposes a system and method for recovering inertial sensor data on a microcontroller. The system comprises a data acquisition and preprocessing unit (100), an event monitoring unit (200), a predictive inference module (300), an asynchronous predictive buffer (400), and a real-time state estimator (500). Angular velocity and acceleration data are dimensionally reduced and stored in the pre-incidence buffer. In response to the detection of a saturated gyroscope, the predictive inference module (300) generates a sequence of multiple virtual orbital compensation values ​​in a single inference and loads this sequence into the asynchronous predictive buffer (400) to sequentially feed the Extended Kalman Filter (EKF). In response to the detection of an acceleration difference event exceeding the impulse threshold, a noise penalty factor is used to adjust the accelerometer's measured noise matrix. As the gyroscope recovers, the recovered data is gradually blended with the real data to minimize signal jumps.

Images