Toward a Unified Bayesian Approach to Hybrid Metric--Topological SLAM

Abstract

This paper introduces a new approach to simultaneous localization and mapping (SLAM) that pursues robustness and accuracy in large-scale environments. Like most successful works on SLAM, we use Bayesian filtering to provide a probabilistic estimation that can cope with uncertainty in the measurements, the robot pose, and the map. Our approach is based on the reconstruction of the robot path in a hybrid discrete-continuous state space, which naturally combines metric and topological maps. There are two fundamental characteristics that set this paper apart from previous ones: 1) the use of a unified Bayesian inference approach both for the metrical and the topological parts of the problem and 2) the analytical formulation of belief distributions over hybrid maps, which allows us to maintain the spatial uncertainty in large spaces more accurately and efficiently than in previous works. We also describe a practical implementation that aims for real-time operation. Our ideas have been validated by promising experimental results in large environments (up to 30 000 m 2 , a 2 km robot path) with multiple nested loops, which could hardly be managed appropriately by other approaches.Index Terms-Bayesian filtering, hybrid metric-topological (HMT) maps, loop closure, mobile robots, Rao-Blackwellized particle filters (RBPFs), simultaneous localization and mapping (SLAM), topological maps. NOMENCLATURE m HMT map (an annotated graph). a M Local metric map for the area a. b a ∆ Coordinate origin of area b relative to that of area a. s t Robot HMT pose at time step t. u t , o t Robot actions and hybrid observations at time step t. s t , u t , o t Sequences of robot poses, actions, and observations for time steps 1 to t. i s t , i u t , i o t A convenient way of referencing the robot poses, actions, and observations grouped into the area i such that the first elements are given for t = 0. ). This material includes two videos (HMT-SLAM malaga.avi; HMT-SLAM edmonton.avi) demonstrating the application of University of Málaga's Hybrid Metric-Topological (HMT) SLAM method to the Málaga Campus dataset and Edmonton dataset, respectively. The first video shows how the robot closes a number of large loops in a large scale (2 km path), nested loop environment. The second video shows how the robot maps a midsized environment with one loop. The first video is of size 28.8 MB while the second is of 19.3 MB. Contact jlblanco@ctima.uma.es for further questions about this work.Color versions of one or more of the figures in this paper are available online at Sequences of all the corresponding variables up to time step t. ψ t , z t Area-dependant and metric observations, respectively. γ t , x t Topological and metric parts of s t at time step t, respectively. γ t Topological path of the robot up to time step t. Υ t Set of all known areas at time step t. s[k ] t kth particle at time step t for the robot HMT pose. ω[k ] t Importance weight of the kth particle at time step t.

Real-time topometric localization

“…Tomatis et al [16], Kouzoubov and D. Austin [10], Bosse et al [6], and Blanco et al [5] used hybrid approaches to connect local metric submaps using high level topological maps. In all cases, the fusion aims at segmenting metric maps represented by topological nodes in order to organize and identify submap relations and loop closures.…”

“…Most existing visual localization approaches fall into two categories: metric or topological. Metric localization produces exact measurements of the observer's pose or position on a map, typically using triangulation [14] or alignment [5]. Topological localization estimates the observer's location qualitatively from a finite set of possible positions.…”

Segmented SLAM in three‐dimensional environments

“…They use a Markov-chain Monte Carlo (MCMC) approach to estimate the distribution over possible topologies by sampling from the space of partitions of landmark measurements. Similarly, the hybrid metric-topological SLAM of Blanco, Fernández-Madrigal, and González (2008) uses the particles of an RBPF to sample topology between evidence-grid metric submaps while using Kalman filters to estimate the transformations between maps.…”

“…See Figure 6. Blanco et al (2008) has a more exact segmentation method using graph cuts but then needs to reconstruct the maps, a slow procedure in two dimensions and an intractable one in three dimensions. We take the penalty of suboptimal segmentation in exchange for real-time speed.…”

Map segmentation for simultaneous localization and mapping in ruins

“…Among them, the hierarchical SLAM based on the hybrid map representation has received a great attention. [2,3] This algorithm combines advantages of the interactivity of topological maps and the accuracy of metric maps. Moreover, because of the composite map representation, the algorithm is more suitable for the human-robot cooperation in rescue CONTACT mhwang@sia.cn and can deal with the large-scale and complex ruins.…”