Abstract
In this paper, we proposed a calibration method for an IMU already attached to a mobile robot by using an long short-term memory (LSTM) neural network. The LSTM network takes IMU and lidar measurements as inputs and outputs compensated IMU measurements. The proposed method offers the advantage of compensating for deterministic errors of the IMU, including settling misalignment errors, while maintaining the IMU attached to the mobile robot. In the LSTM training process, true values of the IMU measurements are required. To obtain these values, we used an external camera to capture a marker on the mobile robot. By numerically differentiating the position and attitude calculated from the captured image, we acquired the true acceleration and angular velocity. We verified the performance of the proposed IMU error compensation method through experiments by using a mobile robot in an indoor environment.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
