Summary
This paper presents the results of experiments in applying a spiking neural network to control the locomotion of a simulated biped robot. The neural model used in simulations was developed to allow for an analytic solution to a neuron fire time, while maintaining a non-instant post-synaptic potential rise time. The synaptic weights and delays were tuned using an evolution strategy. Simulation experiments demonstrate that within about seven thousand generations the biped is able to acquire a dynamic walk which allows it to walk upright for several metres.
See the full content of this document
Extract
Simulated 3d Biped Walking with an Evolution-Strategy Tuned Spiking Neural Network
(ProQuest: ... denotes formulae omitted.)
1. IntroductionControlling the walk of a 3D biped with ? oint- contact feet is a difficult control problem, as the robot is highly unstable, requiring continuous control to maintain an upright walk. In this computer simulation study, experimental results from the application of a spiking neural network to controlling the walk of such a biped are presented.The motivation to use spiking neural networks in this study is to broaden the available set of tools which can be applied to robot control problems from a machine learning perspective. An evolved spiking neural network coupled with a conventional control method has been shown to work when applied to the acrobot swing- up and balance task [15]. In this study, the spiking neural network controller is not augmented with conventional control methods.The field of biped walking control is quite advanced, with many successful solutions implementing sophisticated control methods that require thorough analysis of the dynamics of the particular biped models used, such as virtual constraints [13], or Zero Moment Point preview control [5]. In contrast, biped dy...See the full content of this document
Sponsored links
