Combining PMHT with classifications to perform SLAM

The problem referred to as simultaneous localisation and mapping (SLAM) requires estimation of unknown target locations when the platform location knowledge is unreliable. It is a technique often associated with autonomous platforms that carry a variety of complementary sensors. Besides target detection and platform positional information, these sensors, such as laser ranging and cameras, can often provide perceived classification information that is generally not utilised by the data association algorithm. This paper demonstrates how classification information can be used to assist the data association technique known as the Probabilistic Multi-Hypothesis Tracker (PMHT) when applied to the feature-based SLAM problem. Some example results are given to illustrate the performance improvement that can result from this approach.