[Paper Review] Metric Map Merging using RFID Tags & Topological Information.

Emmanouil Tsardoulias, Aristeidis G. Thallas+1 more|arXiv (Cornell University)|Nov 17, 2017

Robotics and Sensor-Based Localization5 references3 citations

TL;DR

This paper proposes a metric map merging approach for multi-robot systems using RFID tags and topological features to align local occupancy grid maps (OGMs). It computes approximate rotation via obstacle and RFID pose alignment, estimates translation using the best-localized RFID pair, and refines alignment with ICP, achieving near-perfect map fusion with a mean square error near zero and execution time under 18 seconds per merge.

ABSTRACT

A map merging component is crucial for the proper functionality of a multi-robot system performing exploration, since it provides the means to integrate and distribute the most important information carried by the agents: the explored-covered space and its exact (depending on the SLAM accuracy) morphology. Map merging is a prerequisite for an intelligent multi-robot team aiming to deploy a smart exploration technique. In the current work, a metric map merging approach based on environmental information is proposed, in conjunction with spatially scattered RFID tags localization. This approach is divided into the following parts: the maps approximate rotation calculation via the obstacles poses and localized RFID tags, the translation employing the best localized common RFID tag and finally the transformation refinement using an ICP algorithm.

Motivation & Objective

• Address the challenge of fusing local metric maps from multiple robots in multi-robot exploration systems.
• Enable accurate map merging without relying on odometry or global positioning by leveraging environmental landmarks.
• Overcome the lack of a common reference frame in distributed robotic systems through RFID tags and topological features.
• Improve map alignment precision by combining geometric constraints (obstacle orientation), RFID localization, and iterative refinement.
• Minimize inconsistencies in fused maps due to discrete grid representations through a modified blurring process.

Proposed method

• Estimate the relative rotation between two OGMs using the direction vectors of environmental obstacles assumed to be orthogonal.
• Compute the rotation quaternion via RANSAC-based topological alignment of common RFID tag poses across maps.
• Determine the translation using the most accurately localized common RFID tag pair between maps.
• Refine the initial transformation using the Iterative Closest Point (ICP) algorithm to achieve high-precision alignment.
• Apply a modified blurring process to resolve inconsistencies introduced by discrete grid representation after transformation.
• Use a probabilistic Bayesian filter for RFID tag localization, assuming uniform omnidirectional antenna response.

Experimental results

Research questions

• RQ1 Can RFID tags and topological features be effectively used to establish a common reference frame for metric map merging in multi-robot systems?
• RQ2 How accurately can the relative rotation between two local OGMs be estimated using obstacle geometry and RFID tag poses?
• RQ3 What is the impact of using the best-localized RFID pair on translation estimation accuracy in map merging?
• RQ4 To what extent does ICP refinement improve map alignment accuracy compared to geometric and RFID-based methods alone?
• RQ5 How does the proposed method perform in terms of execution time and error metrics (e.g., MSE) under realistic conditions?

Key findings

• The first method—using only RFID tag triplets for transformation—produced highly erroneous alignments with large mean square error (MSE).
• The second method, combining obstacle direction vectors and the best-localized RFID pair, achieved a significant reduction in MSE and visually aligned obstacles correctly.
• The third method, which included ICP refinement after initial transformation, achieved near-zero MSE, indicating almost perfect map alignment.
• Execution time increased with precision: the first method took ~0.02 ms, the second ~10 seconds (due to RANSAC), and the third ~18 seconds (due to ICP).
• The algorithm demonstrated high angular accuracy with a mean rotation deviation of ±1.1196° (95.45% within ±2σ), indicating strong consistency.
• Despite longer execution times, the ICP-enhanced method was deemed preferable due to its high precision, justifying computational cost in off-line map fusion.