Simultaneous Localization and Mapping

Description

Simultaneous Localization and Mapping (SLAM) is a foundational algorithm in robotics and augmented reality that addresses the 'chicken-and-egg' problem of navigation: to map an environment, you need to know your location, and to know your location, you need a map. 3GPP has standardized support for SLAM to enable advanced location-based services in 5G and beyond, particularly for verticals like industrial IoT, autonomous systems, and extended reality (XR). The 3GPP system provides the necessary data and reference frameworks to assist user equipment (UE) or network-based entities in performing SLAM more accurately and efficiently.

Technically, SLAM works by fusing data from multiple sensors. A device uses onboard sensors (e.g., cameras, LiDAR, inertial measurement units - IMUs) to perceive features in the environment (landmarks). As the device moves, it observes these landmarks from different perspectives. The SLAM algorithm, often based on probabilistic frameworks like Kalman filters or particle filters, iteratively estimates two coupled variables: the device's pose (position and orientation) and the 3D positions of all observed landmarks, building the map incrementally. The challenge is managing uncertainty and data association—correctly identifying which new observation corresponds to which previously mapped landmark.

3GPP's role, as defined in specs like TR 26.928, is to provide enhancements that aid the SLAM process. This includes delivering high-accuracy absolute positioning data (e.g., from LTE/5G positioning methods like OTDOA or RTT) to correct the drift inherent in sensor-only SLAM. The network can also provide pre-existing partial maps or semantic information (e.g., floor plans) as a prior, or offload computationally intensive parts of the SLAM processing to edge cloud servers. This network-assisted SLAM improves reliability, reduces the computational burden on the UE, and enables collaborative mapping where multiple devices contribute to a shared map.

Purpose & Motivation

SLAM technology exists to enable autonomous operation and advanced contextual awareness in environments where prior detailed maps are unavailable or dynamic. For robots, drones, and AR/VR devices, GPS is often unavailable indoors or insufficiently precise. Pure inertial navigation (IMU) drifts rapidly over time. SLAM solves this by creating a map on-the-fly, allowing for sustained navigation and interaction.

3GPP's standardization of SLAM support, starting in Rel-16, was motivated by the needs of new 5G verticals. Industrial automation requires autonomous guided vehicles (AGVs) to navigate factories; augmented reality needs persistent world-locked content. Previous 3GPP positioning services (e.g., A-GNSS, OTDOA) provided absolute location but lacked the dense, relative environmental understanding needed for these applications. Standalone device-based SLAM has limitations in scale, consistency across devices, and accuracy over long durations.

By integrating SLAM with the cellular network, 3GPP addresses these limitations. The network provides absolute anchor points to eliminate SLAM's cumulative drift, enables multi-user map sharing and persistence, and allows for computationally efficient split-processing architectures. This creates a scalable, reliable positioning and mapping service that is a key enabler for the metaverse, Industry 4.0, and autonomous systems.

Key Features

• Simultaneous estimation of device pose and environmental map in real-time
• Fusion of visual, LiDAR, inertial, and cellular radio measurements
• Network provision of absolute positioning anchors to correct sensor drift
• Support for collaborative mapping across multiple user devices
• Cloud/edge offloading of computationally intensive map optimization
• Integration with 3GPP positioning architecture (LPP, NRPPa) for assisted data

Evolution Across Releases

Defining Specifications