RescueRobots Freiburg - RoboCupRescue Robot League Team

Results Osaka 2005

Our robots were able to perform both Simultaneous Localization And Mapping (SLAM) and autonomous victim identification during the final of the "Autonomy Competition". Finally, the judge accepted the detection of two victims. The difficulty of the arena, for example, can be seen in the following video: autonomousCompetition.mpg. The final of the autonomous mission can be seen in the following video: autonomyMission.mpg

1. Autonomous Mapping

The following animation shows the mapping process of the robot during the autonomy mission. In contrast to teleoperation missions, maps have to be build 100% autonomously, i.e. without any hand alignment or correction by a human. As you can see, the map has been reliably constructed, although neither odometry nor readings from the laser range finder (LRF) were reliable. The floor of the arena, which was sometimes made of metal and sometimes made of stone, was covered with many newspapers. Furthermore, some of the arena walls were made of glass or pallets, which sometimes appeared as "invisible" to the LRF.

Autonomous SLAM

2. Autonomous Victim Identification

The following series of images has been printed out by the system during the competition. Each time the system detected a victim, the robot stoped automatically and the operator received a message on the screen. If the judge considers the victim detection as invalid, the robot has to wait for 30 seconds. This is the case if the judge is not convinced by the information given by the system.

The images on the left hand side show the Graphical User Interface (GUI) provided to the judge. It shows the thermo image (left side) and the view of a web cam mounted on the robot (lower right corner). The images on the right hand side show the map with victim locations, which have been generated autonomously by the system so far. The red arrow indicates the location of the found victim (which is shown by the GUI).

Victim 1: valid