Dynamically balanced obstacle crossing gait generation of a biped robot using neural networks

Abstract

The gait generation problem of a biped robot while crossing the obstacle is quite difficult in nature due to its inherent complexity. In the present study, two cases, such as landing the foot on the obstacle and placing the foot on the other side of the obstacle while crossing it are considered. During gait generation, the hip and swing foot are assumed to follow straight line and cubic polynomial trajectories, respectively. The gaits related to the lower limbs and trunk are generated after utilising the concept of inverse kinematics and NN-based gait planner. Further, two population-based optimisation algorithms, such as genetic algorithm (GA) and differential evolution (DE) techniques are used to optimise the architecture of NN. Moreover, the performances of the developed GA trained NN (GA-NN) and DE trained NN (DE-NN) gait planners are compared among themselves and with that of the analytical method available in the literature.