Autonomous Direct Calibration of an Inertial Measurement Unit

Sensor calibration is an important step in obtaining useful measurements for an autonomous vehicle. Sensor fusion, in particular, relies on the proper calibration of sensors. Autonomous vehicles are generally designed with a fixed sensor suite. However, this limits the placement and usage of the sensors. Additionally, a manual calibration routine is required before the vehicle can be used. This calibration routine needs to be performed by a set of trained experts to a high degree of precision that requires time and specialized instruments. To enable dynamic reconfiguration of sensors, this work proposes a novel online method to autonomously calibrate an inertial measurement unit (IMU) directly to the vehicle frame. Once the self-calibration has been performed, the other sensors on the vehicle can be calibrated relative to the IMU. The self-calibration is conducted in a two-stage process. First, a Gaussian Radial Basis Function Neural Network is used to emulate an IMU for an arbitrary fixed control point on the vehicle. Then, a constrained maximum likelihood search method performs an IMU-to-IMU calibration between an IMU placed on the body of the vehicle, and the emulated IMU at the control point. The IMU emulation method obtains high-fidelity acceleration estimates on both simulated and experimental data sets. The maximum likelihood search method obtains sensor position estimates within 2 mm of the true sensor location in every direction and within 0.1 degrees of the true sensor orientation for a battery of tests in simulation.