A Neural Network based Approach for Planning Human-like Locomotion in Biped Robot

P. Di Giamberardino and R. Saccone (Italy)


Neural Networks, Motion Planning, Robotics


In the paper a reference generator for the motion of biped robots is proposed. A seven links equivalent robotic structure has been considered. The solution, based on the replication of the human motion, acquired by some measurements, is composed by two levels: a neural network scheme which generates some of the Fourier coefficients of the series expansion of the measured motion, and a function generator which produces the function sum of the Fourier terms generated by the neural network. This scheme, replicated for each of the six joints, is used as a reference trajectory generator to drive a full robot walk.

Important Links:

Go Back