Members-Internal

Study of Self-repair Spiking Neutral Networks for Robot Control

ASL Wiki

Motivation and Background

In recent years, the advance of software-based AI (artificial intelligence) has increased the demand for realization of power-efficient and fast AI computing hardware devices and systems overcoming the bottleneck of the conventional Von Neuman computing style.The brain-inspired cognitive computing in both learning and inference tasks. In SNN, besides the pattern code, the time-related factors,eg.spiking rate (frequency), spiking rank and spiking intervals, are generally used to present the information. The human brain is fault tolerant and continuously adapting a changing environment.Spiking Neural network have many models such as Integrate and Fire(IF), Leaky Integrate and Fire(LIF),Hodgkin Huxly Model(HH) and so on. In some Application, sometimes occurs some faults such as sensor data noises in Unmanned Aerial Vehicles. The features of SNN can be found in repairing such faults.

...to continued

Research Goal

This study provides a decision of efficient fault tolerant algorithm and experiment how SNN repair itself from faults using a Robot car.

...to continued

RPS and RPR seminars

References


トップ   新規 一覧 検索 最終更新   ヘルプ   最終更新のRSS