Kinematics and dynamics of reconfigurable modular robots

We analyze the architecture of modular robots. A total of nine types of connecting forms and three types of joint forms have been identified. The coordinate system associated with each form is established and the geometrical relationships are derived. For achieving a better performance, neural network approach has explored to design the modular robot and determined the sizes of all modules. For the purpose of automatically generating kinematics,a new approach has been adopted by a series of elemental matrixes multiplication. With 4 sub-problems the inverse kinematics can be generated automatically. The dynamic modelling is generated automatically with a compensative recursive method.