Magnetometer Calibration -- Before utilizing the magnetometer data, calibration was necessary to eliminate errors. For this project, only non-time-varying errors—specifically hard-iron and soft-iron distortions—were addressed. Magnetometer readings were collected by driving the car in a circular path, and a linear transformation was applied to calibrate the sensor.

​​

​

​

​

​

​

​

​

​

​

​​

​

 

Sensor Data Fusion -- Yaw estimates from the magnetometer and gyroscope were fused using a custom complementary filter. The fused data demonstrated satisfactory accuracy when compared to the yaw estimates from the IMU, which utilized a Kalman filter.

​​

​

​

​

​

​

​

​

​

​

​

​​

 

Velocity Estimation -- Forward velocity was estimated by integrating acceleration data from the gyroscope. To address inherent biases in the gyroscope, the data was corrected by identifying stationary periods in the car's motion. Using this corrected velocity and the yaw estimates obtained through sensor fusion, dead reckoning was performed to estimate the car's trajectory. The resulting trajectory aligned closely with GPS data during the initial 15 seconds but began to drift over time.