This paper is focused on path planning in environments modelled using continuous probabilistic maps, in particular, maps where obstacles are modelled using the sum of Gaussian distributions. Potential field and roadmap based methods are suitable for these type of maps, but they have some disadvantages. In order to attenuate the disadvantages of the previous methods, a new method has been proposed which is a mixture of them. It performs path planning based on a potential field taking into account a roadmap as a source of potential. Besides, some experiments have been done in order to compare the performance of them.
Keywords: Roadmap, potential field, probabilistic maps, path planning, Gaussian distribution, neural network
JCR Impact Factor and WoS quartile: 0.254 (2004); 2.646 - Q2 (2020)
DOI reference: 10.1023/B:JINT.0000034339.13257.e6
Published on paper: May 2004.