Automatic centralized controller design for modular and reconfigurable robot manipulators

We address the problem of controlling modular robot manipulators. The challenge of modular-robot control is that the overall system dynamics are unknown due to its flexible composition from given modules. Most previous work has faced this problem by designing decentralized controllers. Simple decentralized controllers do not guarantee global asymptotic stability without knowledge of the overall system dynamics and alternative versions involving communication with neighboring modules result in complicated control concepts. Our approach is completely different: we store parameters regarding the dynamics and kinematics of each module and a unique identification number in itself. After finishing the assembly of the modules, the parameters are gathered in a central controller, which also detects the configuration using the identification numbers. Our centralized controller uses this information to synthesize model-based control laws on-the-fly as if the full system dynamics are known beforehand. We introduce a novel and compact notation to automate this procedure and to generalize the derivation of the kinematic and dynamic model for heterogeneous modules. Finally, a possible application is shown using simulations.