problem about improving the precision of odometry?
Recently，I am considering a problem about using rtabmap in the environment which has dynamic objects. Dynamic objects will affect the precision of odometry , so could I consider moving dynamic objects, whether the strategy is effective?
Re: problem about improving the precision of odometry?
Visual odometry in dynamic environments is challenging. RTAB-Map's visual odometry assumes a static environment, though few moving objects may not affect the odometry (if visual features extracted on these objects are set as outliers).
You can create your own odometry algorithm and feed RTAB-Map (on ROS) with its output. I'm not aware of visual odometry strategies to ignore or use moving objects for transformation estimation. Detecting which features are moving and those who are static could be a good start for such algorithm.