CLAUS is robot which may be operated as a walking and as a climbing robot which adheres to the wall by passive suction cups, just like our climbing robot DEXTER. It hence also needs no energy for adhesion. It has four legs, each with three degrees of freedom.

The robot WALTER (WALking TEst Robot) is a quadruped walking robot with simplicity and robustness as its ultimate design goals. It has two degrees of freedom for each leg and an additional servo for panning the batteries. To allow an active control of the center-of-gravity.

Walking and climbing robots quickly get too complicated due to their large number of degrees of freedom which makes the mechanical design as well as control complex and error-prone. On the other hand, wheeled mobile robots only have a limited need of coordination at lower control levels. Both limitations hamper basic investigations on emergent and self-organizing behaviour.

In order to concentrate on basic effects of emergent behaviour, learning and self-organization, a type of autonomous mobile robots was developed which is as simple as possible, but yet robust and scaleable. Scalability is of importance here for testing the generalisability of methods and results. The family of the creeping robots CARL (Creeping Autonomous Robot for Learning demonstrations) hence is scaleable in two directions, namely the number of joints and the number of dimensions these joints may operate in.

The CARL robots move forward in a cyclic, wave-like manner, just like a pine caterpillar. The underlying control behaviour is described only by simple interacting, local rules and yields emergent, global system behaviour. First steps were carried out to make the robot learn, i.e. CARL learned to optimize the step-width achieved by one cycle. Future learning addresses adaptation to the ground surface and to faults.

PACO (Pendulum-cart with Adaptive COntrol) is an omnidrive-platform capable of swinging-up and balancing a pendulum. The aim is to investigate self-optimization strategies under real-life conditions such as slippage, vibration and other disturbances which are introduced by steering and turning the platform along arbitrary trajectories. The rich set of sensors allows for detecting and fighting anomalies. This makes PACO well suited for investigations on trust management as well. PACO is thus an intuitive platform for investigations on self-optimization (SHARCS), on trust management (TIGERS) and on their combination.

FRANC (Fast Running Adaptively Navigating Car) builds on a standard 4WD RC-chassis (1:8 scale). For investigations in our lab, it currently features an electric gear drive. Optionally it can be equipped with a combustion engine which allows for high speed (50 km/h) outdoor operation. Its main scope is to act as a workhorse for trust management investigations (TIGERS). FRANC thus allows capturing its movement by odometry, by IMU (3 DoF accelerator +3 DoF gyro) and by laser scanner. Additional sensors monitor the robot's health state (e.g. motor currents and temperature as well as accumulator state).