SLAM with a Heterogeneous Prior Map

In this paper we propose a method to apply prior geometric information to Kalman Filter- based SLAM. The prior information takes the form of a map which is used in the SLAM process, and need not have the same form as the SLAM map. We validate the method using simulations based on a large urban environment, and show that the prior map significantly improves the performance and consistency of SLAM. We show that this improvement extends to situations where 60 % of the beacons are close to, but not part of a feature in the prior map.