Graph-Based Simultaneous Localization and Mapping

Graph-based simultaneous localisation and mapping (SLAM) is an approach that formulates the challenge of constructing a map of an unknown environment while simultaneously estimating the position of a mobile agent as a sparse graph optimisation problem. In this representation, nodes correspond to robot poses or landmarks, and edges encode spatial constraints derived from sensor measurements such as odometry, laser scans or visual features. The front end of the system handles data association and the detection of loop closures, while the back end performs nonlinear least-squares optimisation over the entire graph or in an incremental fashion. Advances in sparse linear algebra, incremental solvers and robust cost functions have greatly improved both the scalability and the consistency of graph-based SLAM, enabling real-time performance on resource-constrained platforms. This framework is highly adaptable and underpins many applications, from autonomous vehicles and aerial drones to augmented reality and planetary rovers. Its modularity allows the integration of additional sensing modalities—such as inertial measurement units or magnetic field sensors—through the introduction of new constraint types. Recent work has focused on enhancing robustness to outliers, reducing computational overhead and extending the method to multi-agent and large-scale environments without sacrificing global accuracy.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Contemporary studies have proposed a factor-graph model incorporating local constraints to fuse magnetic field readings with pedestrian dead-reckoning data, casting the navigation problem as a hard-constrained optimisation and achieving sub-metre accuracy in complex trajectories. Other work has demonstrated a semi-automated framework for town-scale 3D mapping by aligning pose graphs with building footprints from public maps, drastically reducing manual intervention while yielding globally consistent city models. Foundational contributions include an incremental pose-graph optimisation algorithm that eschews graph marginalisation by estimating relative poses directly, improving robustness to loop-closure outliers and delivering stable performance on real-world datasets.

Graph-Based Simultaneous Localization and Mapping publication trend

The graph below shows the total number of articles in graph-based simultaneous localization and mapping across all publications each year (not limited to Nature Index journals).

Technical terms

Graph-based SLAM: A formulation of SLAM in which robot poses and landmarks are represented as nodes in a graph, with edges encoding spatial constraints from sensor measurements.

Factor graph: A bipartite graph representation separating variable nodes (e.g. poses) and factor nodes (constraints), facilitating efficient optimisation via message passing or sparse solvers.

Pose graph: A specialised graph where nodes represent the estimated position and orientation of the sensor platform, and edges represent relative-pose measurements between these nodes.

Loop closure: The detection and incorporation of a return to a previously visited area, providing constraints that correct accumulated drift in the trajectory estimate.

Optimisation: The process of minimising a global cost function—typically a sum of squared residuals—to find the most consistent set of node estimates given the measurement constraints.

References

1. Factor Graph with Local Constraints: A Magnetic Field/Pedestrian Dead Reckoning Integrated Navigation Method Based on a Constrained Factor Graph. Electronics (2023).
2. Semi-Automatic Town-Scale 3D Mapping Using Building Information From Publicly Available Maps. IEEE Access (2022).
3. Incremental 3-D pose graph optimization for SLAM algorithm without marginalization. International Journal of Advanced Robotic Systems (2020).