A Random Set Approach to SLAM
Authors: John Mullane, Ba-Nu Vo, Martin Adams, Wijerupage Sarha Wijesoma.
This presentation offers an alternative formulation for the Bayesian feature-based simultaneous localisation and mapping (SLAM) problem, using a random finite set approach. For a feature based map, SLAM requires the joint estimate of the vehicle location and the map. In most feature based SLAM algorithms, so-called “feature management” algorithms as well as data association hypotheses and extended Kalman filters are used to generate the joint posterior estimate. Current vector-valued formulations require the data association problem, and map management issues, to be solved prior to the Bayesian state update. This is because the map estimates and measurements are rigidly ordered in a finite-vector-valued map state.
This presentation, however, shows a recursive filtering algorithm which jointly propagates both the estimate of the number of landmarks, their corresponding states, and the vehicle pose state, without the need for explicit feature management and data association algorithms. This is possible since the map estimates and measurements are represented by finite-valued-sets, in which no distinct order is assumed.
Using a random finite-set-valued joint vehicle-map state and set-valued measurements, the first order statistic of the set, called the intensity, is propagated through a probability hypothesis density (PHD) filter, from which estimates of the map and vehicle can be jointly extracted. An extended-Kalman Gaussian Mixture implementation of the recursion is then tested for both feature-based robotic mapping (known location) and SLAM. Results from the experiments show improved performance for the proposed SLAM framework in environments of high spurious measurements.