The purpose of the research addressed in this paper is to develop a real time neural control algorithm for the balance of a biped robot. Our approach is based on dynamic recurrent neural networks and dynamic backpropagation through time algorithm. The neural architecture and its learning process are validated on the control of the ROBIAN biped torso. The neural controller described is trained to compensate, by the torso’s joint motions, applied external perturbations. The algorithm is embedded in the real time electronic unit of the robot and online learning is achieved. The learning behavior and the control performances are the preliminary results presented in this paper. These experimental results show the ability and efficiency of the proposed approach.

This paper concerns the dynamic simulation of an articulated structure, modelling the human upper limb. Our aim is to carry out a highly realistic simulation of the anthropomorphic model of the hand and forearm system. In order to establish the equations of motion of the studied anthropomorphic mechanism, we introduce an uncommon formulation in the mechanics of multibody system: the Korenev's formalism. The use of such formalism allows us to realize this simulation in real time. By real time simulation, we mean that simulated movement has to done at the same rates those of the human. Indeed, the computation time for solving the dynamic forward and inverse models takes less than 0.1 ms for each step. The paper presents this formalism and the model, which structural and functional properties permit the simulation of free movements. The proposed model has 27 degrees of freedom (dofs), which are split into 23 dofs for the hand, 2 dofs for the flexion-extension, and the abduction-adduction of the wrist and 2 dofs for the pronation-supination and the flexion-extension of the forearm. This model is based on bio-mimetic approach, which attempts to simulate the most realistic behavior, with a long term aim of designing an anthropomorphic manipulator robot. Another aim concerns the development of highly realistic simulation platform of virtual manikin animation including not only kinematics aspect but also the dynamic ones.

In this paper, we present the 3D dynamic simulation of walking gait of biped Robian II virtual manikin. The biped has 16 degrees of freedom (DOFs). Initially, a biomimetic approach is used to model a humanoid biped having 25-DOFs based on common European male (75 kg, 1.78 m). Using human being motion recording, foot/ground contact model and inverse kinematics, a 3D dynamic simulation of this humanoid is carried out. Scale factorization is used in order to reach Robian II weight and height. A 3D dynamic simulation of the Robian size humanoid gives the effort wrench exerted by the torso on the lower limbs. An analysis of the six components of this wrench shows the existence of two coupling relations. A study of four DOFs mechanisms based on the general state equation formalism leads us to an interesting result. Indeed, four DOFs are necessary and sufficient to emulate the dynamic effects. A RPPP mechanism is presented in order to replace Robian's upper part. Results of 3D simulation of the 16-DOFs resulting biped are presented. The ZMP control algorithm is used to ensure dynamic stability of the biped during walking gait.