Activity Number:
|
354
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 14, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Bayesian Stat. Sciences*
|
Abstract - #301474 |
Title:
|
Bayesian Hierarchical Spatial Models and Monte Carlo Analysis of the Regular Minefield Patterns in Robotic Landmine Search
|
Author(s):
|
Yangang Zhang*+ and Mark Schervish and Ercan Acar and Howie Choset
|
Affiliation(s):
|
Carnegie Mellon University and Carnegie Mellon University and Carnegie Mellon University and Carnegie Mellon University
|
Address:
|
BH132, 5000 Forbes Ave., Pittsburgh, Pennsylvania, 15213, USA
|
Keywords:
|
Bayesian Hierarchical Model ; Markov Chain Monte Carlo ; Minefield ; Posterior Distribution ; Regular Pattern ; Hasting Algorithm
|
Abstract:
|
We consider the problem of detecting features in spatial point processes in the presence of regular pattern. Our application is to recognize the regular pattern of the minefield based on the point locations of the mines detected by a robot covering a sample area of the minefield. Our solution is to apply a Bayesian hierarchical approach to model the detected mine location data and use Markov Chain Monte Carlo (MCMC) simulation technique to analyze the posterior distribution to recognize the underlining pattern. We solve the difficult computational inefficiency problem in MCMC simulation by developing an algorithm to locate all possible local posterior maximums, which dramatically improves the efficiency of the MCMC simulation algorithm. Therefore, it is feasible to implement our method on a mobile robot in the real time. Our approach allows the assessment of the confidence associated with the ultimate recognized pattern. In this paper, we present our method to recognize a row-shift grid pattern. In this example, the method recognizes the true pattern efficiently. Our method can also be used to detect other general regular patterns.
|