Greater production and improved safety in the mining industry can be enhanced by the use of automated vehicles. This paper presents results in applying Learning from Demonstration (LfD) to a laboratory semi-automated mine inspection robot following a path through a simulated mine. Three methods, Gaussian Mixture Model (GMM), Continuous Hidden Markov Model (CHMM), and Discrete Hidden Markov Model (DHMM) were used to implement the LfD and a comparison of the implementation results is presented. The results from the different models were then used to implement a novel, optimised path decomposition technique that may be suitable for possible robot use within an underground mine.
|Conference||Advances in Artificial Intelligence|
|Period||30/11/15 → 05/12/15|