A Cooperative Unsupervised Connectionist Model to Identify the Optimal Conditions of a Pneumatic Drill

Abstract

A novel connectionist method to feature selection is proposed in this paper to identify the optimal conditions to perform drilling tasks. The aim is to extract information from complex high dimensional data sets. The model used is based on a family of cost functions which maximizes the likelihood of identifying a specific distribution in a data set. It employs lateral connections derived from the Rectified Gaussian Distribution to enforce a more sparse representation in each weight vector. The data investigated is obtained from the sensors allocated in a robot used to drill and build industrial warehouses. It is hoped that in classifying this data related with the strength, the water volume for refrigerating, speed and time of each sample, it will help in the search of the best conditions to perform the drilling of reinforce concrete slabs. This would produce a great saving for the company which owns the drilling robot.