matlabbe
commented
5 years ago First, I'd put my effort on creating an odometry approach that can work accurately in that kind of environment. Once this odometry approach is created, it can be fed to rtabmap for mapping. Unless nodes in rtabmap have 3D point clouds (e.g., 360deg swipe of the 2d lidar), loop closure and proximity detections may not work to correct the odometry drift.
Hi, I understand how RTAB works but I was wondering whether it would to possible to map an environment, for example, a pipe network(a combination of straight pipe and some t-junctions) which has no features (only the junctions can be considered to be a feature). In such case, wheel odometry data becomes quite valuable but I'm not sure if there a way to either increase the weights of odometry than 2d LiDAR and Kinect or any other configuration which can be used in such environment.