Hybrid Control of Biped Robot Leg

J.K. Rai, R.P. Tewari, V.P. Singh, and D. Chandra (India)


Biped Robot Simulation, Flexion Angle, Gait Cycle, Hybrid Control and Neural Network.


This paper proposes a hybrid control based on online neural network and PD controller for fine control of a biped robot leg. The biped robot leg considered here is having three active joints i.e. hip, knee and ankle. The response of proposed hybrid controller is compared with computed torque controller. The neural network is first trained offline with experimental data of human at three joints namely hip, knee and ankle, for inverse dynamics of leg i.e. joint flexion angles and torque data in sagittal plane. The neural network used here is three layered feed forward neural network. This trained neural network is used online with PD controller in feedback loop. Simulation work is carried out in Matlab 6.5 and Simulink 5.0. The result of simulation shows that the response of hybrid controller is better than conventional computed torque control.

Important Links:

Go Back