Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.


Comparison and Experimental Validation of Stochastic Search Algorithms


Stefan Frei

Master Thesis, FS13 (10176)

The lately developed Markov chain Monte Carlo (MCMC) algorithm for stochastic localization of sources is a tool to solve the problem of optimizing a non convex function in the two-dimensional space. It utilizes autonomous vehicles to estimate the source(s) of the underlying concentration eld. We compare its eciency and robustness to a variety of existing methods with numerical simulations. In addition, their application is studied on an experimental test bed with mobile robots. First, auxiliary software is developed such as a visual robot detection method and a controller for the robots. The performance of the setup with the additional software and is analyzed. Second, we devise a collision avoidance technique for Stochastic Localization and show in simulation and experiments that it maintains its convergence property. All methods are implemented in this environment and compared to the simulation. We found that the signi cance of the results obtained from Stochastic Localization and the compared algorithms depend strongly on tuning and the MCMC methods exhibit in general exhibit a rather slower convergence rate to stationarity. Therefore, di erent criteria were investigated to analyze and rate the applicability of the methods to practical search scenarios.


Type of Publication:

(12)Diploma/Master Thesis

S. M. Huck

File Download:

Request a copy of this publication.
(Uses JavaScript)
% Autogenerated BibTeX entry
@PhdThesis { Xxx:2013:IFA_4556
Permanent link