Comparison of Artificial Neural Networks based on Controllers for Biped Robots

Abstract

This paper provides a brief overview the need for humanoid robots specifically bipedal robots. The advantages and shortcomings of bipedal robots have been highlighted. The paper discusses in detail the various types of controllers that have been applied for the control of humanoid robots. Typical controllers such as slide mode control, active force control, computed torque control etc., require the boundaries of uncertainties to be known, which is very difficult considering the real world terrain. The application of artificial neural networks has thus become a priority in controller design for bipedal robots. Some of the neural network based on the controller design and their comparisons to typical controllers have been surveyed in this paper. The survey presented visibly shows that the application of neural networks results in an exponential increase in accuracy for bipedal robots in comparison to classic approaches.