Abstract:
When using an extended Kalman filter (EKF) in
simultaneous localization and mapping (SLAM) for a mobile
robot with bearing-only measurements, it is crucial to correctly
assign correspondences between measurements and registered
features in the map, otherwise the filter diverges or becomes
inconsistent. Conventional methods based on the Mahalanobis
distance metric may produce data association ambiguities. Its
reliability may further be degraded in bearing-only SLAM due
to the limited amount of information delivered from the sensor.
The data association process is cast here as that of making
a decision based on the sensor measurement as whether to
update the EKF or not. For this, cost functions are applied
taking into account the interferences from other features. The
proposed approach enhances robustness of the data association
and consequently assures the performance of bearing-only
SLAM. Results from simulations and experiments are included
to demonstrate the effectiveness of the method in a typical
indoor scenario.
