Differential Evolution Markov Chain Filter for Global Localization

Abstract

A key challenge for an autonomous mobile robot is to estimate its location according to the avail able information. A particular aspect of this task is the global localization problem. In our previous work, we developed an algorithm based on the Differential Evo lution method that solves this problem in 2D and 3D environments. The robot's pose is represented by a set of possible location estimates weighted by a fitness function. The Markov Chain Monte Carlo algorithms have been successfully applied to multiple fields such as econometrics or computing science. It has been demonstrated that they can be combined with the Dif ferential Evolution method to solve efficiently many optimization problems. In this work, we have com bined both approaches to develop a global localization filter. The algorithm performance has been tested in simulated and real maps. The population requirements have been reduced when compared to the previous version.

Parameter sensitivity analysis and optimization for a satellite‐based evapotranspiration model across multiple sites using Moderate Resolution Imaging Spectroradiometer and flux data

“…In the DE‐MC method, N chains are run in parallel and the proposals are generated on the basis of two randomly selected chains, the difference of which is multiplied by a scaling factor, and added to the current chain: where θ p is the proposed parameter set; θ R 1 and θ R 2 represent the randomly selected chains without replacement from the population θ − i (the population without θ i ); e is drawn from a symmetrical distribution with a small variance compared to that of the target, but with unbounded support, e.g., e ~ N (0, b ) d with b small and d being the parameter dimension; and γ is the scaling factor which can be set to be [ Roberts and Rosenthal , ]. The Metropolis ratio is then used to decide whether to accept or reject the proposals [ Moreno et al ., ].…”

“…where θ p is the proposed parameter set; θ R1 and θ R2 represent the randomly selected chains without replacement from the population θ À i (the population without θ i ); e is drawn from a symmetrical distribution with a small variance compared to that of the target, but with unbounded support, e.g., e~N(0,b) d with b small and d being the parameter dimension; and γ is the scaling factor which can be set to be 2:38= ffiffiffiffiffi ffi 2d p [Roberts and Rosenthal, 2001]. The Metropolis ratio is then used to decide whether to accept or reject the proposals [Moreno et al, 2016].…”

Kullback-Leibler Divergence-Based Differential Evolution Markov Chain Filter for Global Localization of Mobile Robots

“…As can be verified in the experimental results, the method performance is excellent in environments with occlusions, which is a very interesting characteristic that makes this technique a suitable approach for environments with dynamic objects and people. In addition, the population requirements are similar to those presented in [ 11 ].…”

“…He has applied his method to multiple optimization problems. In our recent research, we have designed a GL module based on the DE-MC algorithm [ 11 ]. Its main advantages are the improvement in the robustness (regarding the chances of success in the GL process) and the decrease of the population requirements with respect to the basic version (DE) of the filter.…”

“…In this work, a new version of the filter where the KL divergence-based cost function [ 5 ] is implemented together with the DE-MC GL algorithm [ 11 ] is presented. The main objective is to obtain a GL module that inherits the advantages of both approaches.…”