The use of robots for search and rescue tasks has tremendous potential to mitigate disasters and save lives of both of disaster victims and first responders. Moreover, robots actually deployed for disaster response are neither highly intelligent (i.e. autonomous) nor do they intelligently make best use of a human supervisor in the loop nor do multiple heterogeneous robots work together in an intelligent manner.

Within this thesis, a holistic systems-oriented approach together with a number of developments of key functionalities to increase robot autonomy for rescue robots systems are presented and evaluated. These are usable for a wide range of robotic systems and operation modes, from unmanned ground vehicles for exploration and surveillance to complex high degree of freedom humanoid robotic systems that can be used for remote manipulation in disaster environments and thus, may potentially serve as "avatars" for human response forces supervising them. Importantly, cooperation between one or more human supervisors and such robotic systems is demonstrated, increasing overall reliability and capability.

The presented approaches are evaluated in simulated and real world robot experiments and presented within the scope of some of the most competitive international competitions for rescue and disaster response robots in the world, the the DARPA Robotics Challenge competition and the RoboCup Rescue Robot League, demonstrating their performance beyond laboratory environments and representing an important milestone towards their future use in real disaster environments.