Title: Reliable Real-Time SLAM within Search and Rescue Scenarios

Speaker: Alexander Kleiner (web page)

Affiliation: Foundations of Artificial Intelligence, University of Freiburg, Germany

Abstract:
Search and rescue is a time critical task, i.e. a large terrain has to be explored by multiple humans or robots within a short amount of time.
Rescue teams have to generate a map of the environment which has to be sufficiently accurate for reporting the locations of victims to medics.
Basically, one has to solve in real-time the problem of Simultaneous Localization and Mapping (SLAM), consisting of a continuous state estimation problem and a discrete data association problem. The state estimation problem, on the one hand, is hard due to the extremely unreliable odometry measurements usually found on tracked robots operating within harsh environments. Pose tracking becomes particularly hard while the robot climbs 3D structures, such as pallets, stairs and ramps. The data association problem, on the other hand, i.e. to recognize locations from the data, is challenging due to the unstructured environment, i.e. arbitrarily colored and shaped debris from building collapse and fire.

The research proposed in this talk extends well known methods from robotics towards their usability within the context of Urban Search And Rescue (USAR). The introduced robot system tackles the problem of state estimation with a visual odometry approach combined with laser scan matching. For the solution of the data association problem we introduce a novel approach for the active distribution and recognition of RFID tags utilized by the method from Lu and Milios for building globally consistent maps.

The validity of the system is shown by results from numerous experiment and also by the system's success in achieving the first place during competitions at RoboCup rescue.  (more)