I'm using CMetricMapBuilderRBPF in MRPT to do SLAM, but after the robot makes sharp turns, subsequent LIDAR observations are not correctly aligned with the rest of the map. The robot has a very accurate DSP-3000 gyro, and I'd like to use this orientation data to improve the performance of the SLAM process. What is the best approach to using this gyro data for SLAM within MRPT? Thank you.
There's at present no straight-forward way to do what you want... You could "improve" the odometry readings by using the heading from the gyro, for example.
Or you could just increase the uncertainty of rotations in the noise model of odometry.