A navigation method based on BA-POMDP algorithm

This paper presents a robot navigation method based on hierarchical POMDP and Bayesian Algorithm (BA-POMDP Algorithm) in uncertain environments. A successful and efficient robot navigation method in dynamic environments requires predicting that the uncertainties of state of events as well as obstacles. A novel procedure accounting for both state transition and observation uncertainty in the navigation process is presented. In order to solve the problem in dynamic planning programming that is associated with robot navigation in uncertain environments, we present BA-POMDP algorithm that integrate prediction, estimation and planning while also properly fuse weights by mapping between fusion weights and the immediate environmental configuration. The algorithm is implemented and tested onboard that the elderly companion robot achieves autonomous navigation. Experiments from dynamic scenarios illustrate the effectiveness of the BA-POMDP algorithm.