One of the goals of the RoboCupRescue competition is to provide a standardized testbed for robots that can autonomously navigate in unknown and unstructured environments. This paper describes in detail our contribution to the competition: Robbie. The robot uses an active sensing approach, so the sensors are adjusted or configured before the data is readout, depending on the task or other sensor readings. A map is generated on the fly using the two-dimensional (2-D) distance measurements of a gimbaled laser range finder (LRF). The measured data are fused in an occupancy grid using a combination of scan matching and particle filter. The same LRF generates 3-D scans for the obstacle detection. The result of the obstacle detection and also the current 2-D laser scan are blended into the 2-D map. This new map is the basis for the path planning, where we developed the so-called Exploration Transform that combines the frontier-based exploration with the path transform. While navigating, the planned path is constantly checked against the updated map and replanning is started as soon as an obstacle is detected that blocks the calculated way. The victim detection relies on a handcrafted thermal camera that captures images with a field of view of up to 180° (horizontally). All components are combined using a strictly message-based software architecture. Robbie was used at the RoboCup (German Open and World Championship) in 2007 and 2008, and won the 'Best in Class Autonomy' award in all four competitions.