🔢 mathematics

When Simultaneous Localization and Mapping Meets Wireless Communications: A Survey

This paper surveys the bidirectional integration of Simultaneous Localization and Mapping (SLAM) and wireless communications, highlighting how visual SLAM can leverage RF data for scale resolution while wireless networks benefit from visual odometry, ultimately outlining current techniques, challenges, and the future potential of joint communication and SLAM systems.

Original authors: Konstantinos Gounis, Sotiris A. Tegos, Dimitrios Tyrovolas, Panagiotis D. Diamantoulakis, George K. Karagiannidis

Published 2026-05-22

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Konstantinos Gounis, Sotiris A. Tegos, Dimitrios Tyrovolas, Panagiotis D. Diamantoulakis, George K. Karagiannidis

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to navigate a dark, foggy room while holding a flashlight. You need to know two things at the same time: where you are and what the room looks like. This is the core challenge of SLAM (Simultaneous Localization and Mapping).

For a long time, robots and self-driving cars have tried to solve this using just one tool: their "eyes" (cameras). This is called Visual SLAM. But cameras have limits. If it's too dark, if there's a mirror, or if the camera gets covered in mud, the robot gets lost.

This paper is a big review of what happens when we stop relying only on eyes and start using the robot's "ears" and "skin" to feel the air around it—specifically, wireless signals (like Wi-Fi, 5G, and 6G). The authors argue that combining these two senses creates a super-powered navigation system.

Here is a breakdown of the paper's main ideas using simple analogies:

1. The Two Best Friends: Eyes and Radio Waves

Think of a robot as a person trying to find their way in a maze.

The Eyes (Visual SLAM): These are great at seeing details. They can spot a red door or a specific chair. But they struggle if the lights go out or if the robot moves too fast and the image blurs. Also, cameras often get confused about scale. They might see a toy car and think it's a real car because they can't tell the distance without a reference.
The Radio (Wireless Sensing): This is like having a sonar or a radar. It bounces signals off walls and measures how long they take to come back. It doesn't care if it's dark or if the robot is moving fast. It can tell you exactly how far away a wall is.

The Paper's Big Idea: If you combine them, the radio signals act like a "ruler" for the camera. The radio tells the camera, "That object is 5 meters away," which fixes the camera's confusion about size and distance. Conversely, the camera can help the radio by identifying what the signal is bouncing off (e.g., "That's a car, not a wall"), helping the radio predict where the signal will go next.

2. Passive vs. Active: Listening vs. Dancing

The paper divides this teamwork into two styles:

Passive SLAM (The Listener): The robot just drives along, listening to the radio signals and looking through the camera. It uses the data it naturally receives to build a map. It's like walking through a room and just noticing where the furniture is without touching anything.
Active SLAM (The Dancer): This is where it gets clever. The robot realizes, "I'm not sure where that corner is." So, it deliberately changes its path or points its antenna in a specific direction to get a better signal. It's like a detective who walks around a crime scene specifically to get a better angle on a clue. The robot moves on purpose to make its map more accurate or to get a stronger Wi-Fi signal.

3. The "Digital Twin" and the "Ghost Map"

The paper mentions that we can create a Digital Twin. Imagine a video game version of the real world that updates in real-time.

The robot sends its camera photos and radio data to a powerful computer (in the "cloud" or on the edge).
That computer builds a perfect 3D map of the room.
This map is sent back to the robot instantly.
Even if the robot's camera is blocked by a box, the "Digital Twin" knows the box is there because the radio signals bounced off it. The robot can "see" through the box using the map.

4. The "Smart Surface" (RIS)

The paper talks about a new technology called Reconfigurable Intelligent Surfaces (RIS). Imagine a wall covered in thousands of tiny, smart mirrors.

In the old days, if a signal hit a wall, it just bounced away randomly.
With these smart mirrors, we can program the wall to catch a signal and reflect it exactly where we want it to go.
The paper suggests these walls can actually help robots map the room. The wall acts like a helper, bouncing signals around corners so the robot can "see" areas it couldn't see before.

5. The Challenges (Why isn't this everywhere yet?)

Even though this sounds perfect, the paper points out three big hurdles:

The Foggy Room (Dynamics): Real life is messy. People walk around, lights flicker, and cars move. The robot's map can get confused if things change too fast.
The Tired Brain (Latency & Energy): Processing all this data (images + radio signals) takes a lot of power and time. If the robot takes too long to think, it might crash. If it uses too much battery, it stops working.
The Spy (Security): If a hacker can trick the robot's radio or camera, they can make the robot think a wall is a door, leading to a crash. The paper warns that we need to protect these systems from being tricked.

Summary

This paper is a roadmap for the future of self-driving cars and robots. It says: "Stop trying to navigate with just one sense."

By teaching robots to use their cameras (to see details) and their radio antennas (to measure distance and see through obstacles) together, we can build systems that are safer, smarter, and able to navigate in the dark, the fog, or even when the lights go out. It's about turning the invisible radio waves into a visible map, and using the robot's movement to make the radio signals smarter.

Technical Summary: When Simultaneous Localization and Mapping Meets Wireless Communications: A Survey

Problem Statement
The paper addresses the emerging intersection of Simultaneous Localization and Mapping (SLAM) and wireless communications, a convergence driven by the proliferation of commercial wireless sensing equipment and advancements in intelligent autonomous systems. While SLAM has traditionally relied on visual (RGB, LiDAR) and inertial sensors, and wireless communications have evolved to provide high-frequency, low-latency connectivity, the two domains have largely developed in parallel. The core problem is the lack of a unified framework that exploits the bidirectional impact between these fields. Specifically, there is a need to understand how wireless signals (RF, mmWave, WiFi) can resolve limitations in visual SLAM (such as scale ambiguity and feature loss in texture-less environments) and, conversely, how visual odometry and landmark detection can optimize wireless communication objectives (such as beam alignment and channel modeling). The paper also highlights the gap in existing literature, which often treats wireless localization or SLAM in isolation rather than as a coupled perception-estimation-control system.

Methodology
The authors conducted a comprehensive survey using a three-step literature search strategy involving the Scopus database, a manual review of IEEE Communications Surveys & Tutorials (2018–2025), and a screening of reference lists. The survey is structured around three core thematic axes:

Foundations of SLAM in Networked Autonomous Systems: The paper defines a multilayered architecture (Interaction, Sensing, Estimation, Inference, and Coordination layers) to contextualize how SLAM operates within networked environments. It distinguishes between Passive SLAM (inferring motion and structure from existing data streams without deliberate probing) and Active SLAM (where control inputs are optimized to reduce estimation uncertainty or improve communication metrics).
SLAM Methods and Modality-Based Techniques: The survey categorizes and analyzes mathematical and algorithmic approaches, including probabilistic methods (Bayesian filters, particle filters, factor graphs), graph-based SLAM, and vector-field SLAM. It further reviews sensor modalities such as Inertial, LiDAR, Visual Range Odometry, and multi-agent RSSI-based localization.
Wireless- and 6G-Enabled SLAM: This section examines the integration of wireless technologies into SLAM pipelines. It covers Deep Neural Network (DNN)-based wireless SLAM (Deep Coupling SLAM, Deep Semantic and Spatial Joint Perception), radio map extraction via mmWave sensing, WiFi fingerprinting, device-based sensing (ISAC), and Reconfigurable Intelligent Surface (RIS)-aided SLAM.

Key Contributions

Unified Framework for Bidirectional Impact: The paper synthesizes research threads into a unified framework demonstrating how wireless signals augment the vehicle state vector and how visual information aids wireless channel modeling. It explicitly maps the bidirectional relationship: RF/mmWave signals serve as proxies for scale resolution in monocular V-SLAM, while visual odometry aids in optimal geometric path prediction for wireless channels.
Taxonomy of Active vs. Passive SLAM: The authors provide a detailed distinction between passive SLAM (where communication feedback signals preserve connectivity while conveying geometric information) and active SLAM (where motion control is optimized to reduce landmark uncertainty or improve beam alignment).
Mathematical Formulations: The survey details specific mathematical models, including:
- Epipolar Geometry and Scale Ambiguity: How metric motion inputs from wireless odometry (ToA, AoA, RSS) can resolve the scale factor ( $\lambda_k$ ) in monocular V-SLAM.
- Active Control Formulation: The derivation of control inputs ( $u_k$ ) to minimize the trace of the posterior covariance matrix, effectively linking estimation uncertainty reduction to communication goals like optimal beam steering.
- Vector-Field SLAM: Modeling spatial variations in continuous signals (e.g., WiFi RSS) as vector fields to enable mapping without explicit landmark extraction.
Comparative Analysis: Through Table I and Table II, the paper quantitatively compares its scope against representative surveys (e.g., Zeng et al., Italiano et al., Shastri et al.), demonstrating its unique coverage of joint vision-wireless integration, mathematical modeling, and both passive and active SLAM paradigms.

Results and Findings

Complementarity of Modalities: The survey finds that monocular V-SLAM significantly benefits from RF/mmWave signals for scale ambiguity resolution. Conversely, 5G and beyond networks can leverage visual odometry and landmark detection for predicting channel geometry and optimizing beam alignment.
Emerging Technologies:
- Deep Learning: DNN-based approaches (e.g., DC-SLAM) show superior performance in NLOS environments compared to traditional probabilistic methods, achieving average positioning accuracies of 0.11m in indoor settings.
- RIS-Aided SLAM: The paper discusses "MetaSLAM," where RISs act as distributed nodes providing SLAM services, and "LeRIS" (Light-emitting RIS) which uses optical anchors for localization.
- Radio Tomographic Imaging (RTI): Systems like UBIQMAP demonstrate the ability to map environments beyond line-of-sight (LOS) using RF attenuation, complementing traditional LiDAR/visual SLAM.
Challenges Identified: The paper identifies four primary challenges hindering widespread deployment:
1. Environment Dynamics: Moving objects and time-varying occlusions cause inconsistencies between visual and wireless modalities.
2. Latency: Strict real-time constraints are difficult to meet due to processing delays in high-resolution sensors and communication overhead in networked systems.
3. Energy Efficiency: High computational loads for feature extraction and frequent CSI updates create significant power demands for mobile platforms.
4. Security and Privacy: Networked SLAM systems are vulnerable to adversarial attacks (spoofing, jamming) and raise privacy concerns regarding the sharing of spatial and trajectory data.

Significance and Claims
The paper claims that the intersection of SLAM and wireless communications forms a foundational paradigm for 6G-enabled intelligent robotic and autonomous vehicle systems. It argues that while theoretical and practical advancements are emerging, integrated solutions performing joint communications and localization are currently in their infancy.

The significance of this work lies in its role as a cross-disciplinary reference point. It posits that future networked autonomous systems will require a shift from treating localization and connectivity as separate functions to a unified "Integrated Sensing and Communications" (ISAC) approach. The authors suggest that theoretical and practical advancements in this intersection could add higher-level localization and semantic perception capabilities to RF/mmWave and multi-antenna technologies, ultimately supporting safe navigation, robust state feedback control, and real-time network optimization. The paper modestly concludes that while promising results exist, fully integrated solutions require further research in hetero-modal fusion, distributed estimation, and edge-powered computation.