Building an environment model using depth information

The environment model's volumetric level, where information about free and occupied space is represented explicitly, is considered. At this model level, updating operations use raw sensory data from range sensors or processed data from any stereo or other depth recovery technique. In addition, this type of model can be used directly by path planning and navigation modules as well as for object recognition and manipulation modules. A system that uses the sensor-built model mode, in which there is initially little or no knowledge available about the domain and a model is incrementally constructed using the information provided by a sensor, is demonstrated. This mode of operation makes it possible to deal with unknown or dynamic environments.<
>