top of page

Radio SLAM
1. mmWave/THz Radio SLAM

Millimeter wave (mmWave) signals in fifth-generation (5G) communications can enable accurate positioning for vehicular networks. However, radio-based positioning suffers from multipath signals generated by various objects in the physical environment. Multipaths can be leveraged as a benefit by building a radio map, which comprises the number of objects, object type, and object state. The radio map is utilized to optimize all available signal paths for positioning and mapping, serving as a solution to radio simultaneous localization and mapping (SLAM).
Fig. 1. Radio SLAM scenario with the base station, the vehicle, and four virtual anchors (each corresponding to a vertical wall).

<5G mmWave radio SLAM simulation>

Fig. 2. 5G mmWave cooperative radio SLAM
Fig. 2 illustrates the radio SLAM scenario, where 5G vehicular networks can be utilized to share measurements, radio maps, and location information, thereby enabling the cooperative positioning and mapping. In 5G mmWave cooperative positioning and mapping, three main tasks are as follows:(i) Vehicle positioning: determine the states (position, velocity, heading, and clock bias) of the vehicles;(ii) Environment mapping: estimate the number of objects, as well as each object’s type and location;and (iii) Cooperation: fuse the collected map from vehicles and transmit it to each vehicle. As shown in Fig. 3 below, the mapping of landmarks is achieved quickly by the cooperation between the base station and vehicles.


Fig. 3. Mapping accuracy without the cooperation (left) and with the cooperation (right)

Fig. 4. Radio signal propagation by arbitrarily-shaped objects and flat walls
Most existing radio SLAM works focus on point-like landmarks, such as virtual anchors and scattering points, which can be represented using fixed 3D coordinates. However, real-world environments, particularly in indoor or urban scenarios, contain many irregular reflectors, as shown in Fig. 4. For correctly classifying various types of landmarks, we proposed a novel approach that maps arbitrarily shaped reflectors using a Dirichlet process clustering-based radio SLAM. The proposed technique can detect the types of multiple objects, classify each object to perform the mapping, and utilize this mapping information to improve the positioning accuracy.

Fig. 5. Environment mapping procedure for (a) Active sensing, (b) Passive sensing, (c) Landmark estimation
Conventional radio SLAM frameworks assume a single dominant type of wall reflection, either specular or diffuse, ignoring a mixture of reflection types. Furthermore, single-mode radio SLAM approaches fail to robustly detect or map walls in environments with mixed reflection characteristics. To detect both mirror-like and rough walls robustly, we proposed a hybrid-sensing radio SLAM framework that combines: (i) Active sensing, which is effective for the detection of rough walls; (ii) Passive sensing, which complements mirror-like walls. As shown in Fig. 6, the proposed framework outperforms both active-only and passive-only SLAM in environments with mixed reflection surfaces by integrating data from both sensing modalities into a unified map.

Fig. 6. Mapping performance by the proposed method compared to radio SLAM with only active or passive sensing
[Related publication]
2. AI/ML based radio SLAM
In radio SLAM, reducing information loss and computational complexity is a critical challenge to enable accurate and real-time mapping. As shown in Fig. 7, we propose a machine learning-driven landmark mapping algorithm using channel impulse responses (CIRs). The algorithm leverages CIRs, which preserve the raw characteristics of radio propagation, to enhance mapping performance. Furthermore, it replaces mapping filters in existing radio SLAM algorithms with machine learning to reduce computational complexity and improve efficiency.

Fig. 7. Example of machine learning-driven landmark mapping using CIRs
[Related publication]
bottom of page