Initial population size estimation for a Differential-Evolution-based global localization filter Articles uri icon

publication date

  • July 2014

start page

  • 245

end page

  • 258

issue

  • 3

volume

  • 29

international standard serial number (ISSN)

  • 0826-8185

electronic international standard serial number (EISSN)

  • 1925-7090

abstract

  • The mobile robot global localization problem, which can be defined as the search of the robot's coordinates in a known environment with no initial information, can be solved by diferent methods. The localization filter that is proposed here is an optimization method based on the Differential Evolution algorithm, which is a population-based evolutionary technique that evolves with the time to the true solution. One of the most important challenges of any population-based global localization system is the initialization problem. The initialization problem consists of obtaining an appropriate initial population size for each situation. This number must be large enough to ensure that the global localization algorithm converges to the true pose, and it must be the minimum value that ensures a robust convergence. It is easy to check the first condition, but it is more difficult to deal with the second one. In this paper, we propose a solution that automatically generates an initial population that is based on the environment size and the amount of information contained in the first observation of the mobile robot. The efficiency of the proposed method and its capabilities have been tested in a simulated indoor environment.

keywords

  • genetic algorithms; differerential evolution; initialization; global localization; mobile robots