Sensor Planning for Mobile Robot Localization based on A hierarchical approach using Bayesian network and particle filter
Abstract:
| The paper presents a system which takes a hierarchical approach to solve sensor planning for the global localization of a mobile robot. The higher layer of the system uses a Bayesian network which represents the contextual relation between the geometrical features of local environment, the robot sensing actions, and the global localization beliefs. The layer allows sensor planning by taking into account the trade-off between global localization belief and the sensing cost to generate optimal sensing action. Using the planned efficient sensing actions, the robot can obain a rough localization. The lower layer was implemented using particle filter, which allows mobile robot to perform precise localization. Through the intergration of the two layers, the system attempts to solve the mobile robot localization not only efficiently but also accurately. |