Currently, I need to map a large area that consists of 25 parallel horizontal corridors and 2 vertical main corridors. In the test maps I recorded, I’m satisfied with the localization inside the corridors, but I haven’t been able to solve the long-corridor problem — there are constant drifts. How can I fix this in SlamToolbox?
Hardware-wise, I have a SICK Nanoscan3 LiDAR sensor. The horizontal corridor you see is approximately 80 meters long, and the vertical main corridors are around 130 meters. My robot uses differential drive. It has wheel encoders, but when I included them for SLAM, the drift actually got worse. Therefore, I tried generating odometry from the LiDAR and performing SLAM that way.
Additionally, I have an Adafruit 4646 IMU with the BNO055 chip, but I haven’t really been able to integrate it. I know the BNO055 can be somewhat noisy and not as stable as higher-grade IMUs, but could it still be useful in such a large environment? If so, how should I include it? How should I use its matrix in the EKF? The horizontal corridors I mentioned consist of rows of symmetric machines rather than walls.
Maps and images are shown below.
Is it possible to correct this using interactive mode? I haven’t been very successful. I don’t fully understand how it works.
