Genetic approach for a localisation problem based upon Particle Filters

Localisation, i.e., estimating a robot pose relative to a map of an environment, is one of the most important problems in mobile robotics. The literature offers several possible approaches to deal with such task, and optimal algorithms can be devised relying on particular constraints (linear state equations, Gaussian posterior density, etc.). Here, we propose a preliminary study for an algorithm able to approximate a large number of probability distributions when a map of the environment is available. This work improves the particle filters strategies presented in literature, reducing the number of particles needed to solve the localisation problem. The algorithm relies upon a suitable clustering of the particle set and a genetic approach for the resampling step