Cooperative Localization in Multi-Robot Systems

Cooperative localization refers to the process by which a group of autonomous robots jointly estimate their positions and orientations by sharing sensor data and pose estimates. This strategy overcomes the limitations of individual navigation systems by exploiting inter-robot measurements, communications and relative observations. Approaches span centralised architectures, in which a dedicated node fuses all information, to fully decentralised schemes that distribute computation and reduce communication bottlenecks. Common methodologies include variations of the Kalman filter for sensor fusion, particle filters for non-Gaussian uncertainty and graph-based optimisation to enforce consistency across the team. By pooling range, bearing, inertial and environmental cues, multi-robot systems can maintain accurate pose estimates in GPS-denied or feature-sparse environments such as indoor facilities, underwater domains or urban canyons. This collaboration enhances robustness to individual sensor failures, scales gracefully to larger swarms and enables complex missions such as cooperative mapping, formation control and distributed exploration. Recent advances leverage formal observability analysis, fault detection and consensus algorithms to guarantee bounded estimation errors under communication constraints and dynamic network topologies.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Consistent extended Kalman filter techniques have been adapted to networks of autonomous underwater vehicles operating in leader–follower configurations. By analysing the observability of the linearised estimator and correcting Jacobian terms in the measurement model, researchers achieved synchronised and stable position estimates even under dead-reckoning drift. Simulation studies demonstrate that the consistent filter outperforms the standard formulation in maintaining bounded error growth over extended missions.

In air and ground vehicle networks with high initial uncertainty, multi-hypothesis extended Kalman filters have been employed to initialise pairwise relative poses using range-only measurements. This framework generates multiple candidate hypotheses, selects feasible solutions via observability constraints and then merges them into a joint filter. Hardware experiments on heterogeneous platforms show that this strategy rapidly converges from poor priors and sustains accurate estimates under intermittent connectivity.

Onboard quadrotor platforms, unscented transformation-based collaborative algorithms have been developed to address strong nonlinearities in motion and sensing. By approximating inter-robot correlations in a distributed manner and applying a conservative covariance intersection step for robot–target interactions, these methods yield consistent state estimates of both local agents and shared objectives. Comparative trials indicate superior performance over extended Kalman filter variants, particularly in dynamically changing formations and asynchronous communication scenarios.

Cooperative Localization in Multi-Robot Systems publication trend

The graph below shows the total number of articles in cooperative localization in multi-robot systems across all publications each year (not limited to Nature Index journals).

Technical terms

Cooperative localization: The joint estimation of poses for multiple robots by sharing measurements and beliefs to improve accuracy and resilience.

Decentralised algorithm: A computational scheme in which each robot processes its own data and communicates selectively, without relying on a central coordinator.

Extended Kalman filter (EKF): A nonlinear state estimator that linearises motion and measurement models about current estimates to fuse multiple sensor inputs.

Unscented transformation: A sampling-based method to propagate means and covariances through nonlinear functions more accurately than linearisation techniques.

Observability: A property determining whether the internal state of a system can be inferred from external measurements over time.

References

1. Consistent Extended Kalman Filter-Based Cooperative Localization of Multiple Autonomous Underwater Vehicles. Sensors (2022).
2. Unscented Transformation-Based Multi-Robot Collaborative Self-Localization and Distributed Target Tracking. Applied Sciences (2019).