MapViT: A Two-Stage ViT-Based Framework for Real-Time Radio Quality Map Prediction in Dynamic Environments

TL;DR

MapViT tackles dynamic radio environment awareness for mobile robots by predicting environmental changes and radio quality maps with a two-stage ViT-based framework. It combines self-supervised pre-training on depth maps with a supervised fine-tuning stage and introduces a Geometry Foundation Model to enable data-efficient transfer to geometry-derived modalities. Empirical results show ViT-based pipelines outperform MLP and CNN baselines in both accuracy and runtime, achieving near real-time inference (~1 ms) compared with traditional ray tracing. The work advances digital twin and 6G-enabled multi-modal intelligence by enabling efficient, scalable predictions with limited labeled data.

Abstract

Recent advancements in mobile and wireless networks are unlocking the full potential of robotic autonomy, enabling robots to take advantage of ultra-low latency, high data throughput, and ubiquitous connectivity. However, for robots to navigate and operate seamlessly, efficiently and reliably, they must have an accurate understanding of both their surrounding environment and the quality of radio signals. Achieving this in highly dynamic and ever-changing environments remains a challenging and largely unsolved problem. In this paper, we introduce MapViT, a two-stage Vision Transformer (ViT)-based framework inspired by the success of pre-train and fine-tune paradigm for Large Language Models (LLMs). MapViT is designed to predict both environmental changes and expected radio signal quality. We evaluate the framework using a set of representative Machine Learning (ML) models, analyzing their respective strengths and limitations across different scenarios. Experimental results demonstrate that the proposed two-stage pipeline enables real-time prediction, with the ViT-based implementation achieving a strong balance between accuracy and computational efficiency. This makes MapViT a promising solution for energy- and resource-constrained platforms such as mobile robots. Moreover, the geometry foundation model derived from the self-supervised pre-training stage improves data efficiency and transferability, enabling effective downstream predictions even with limited labeled data. Overall, this work lays the foundation for next-generation digital twin ecosystems, and it paves the way for a new class of ML foundation models driving multi-modal intelligence in future 6G-enabled systems.

Figures (7)

• Figure 1: Architecture Framework.
• Figure 2: The deployment pipeline of MapViT. In Stage 1 (left), a sequence of previous depth maps over a given time window is processed to predict the next depth map. In Stage 2 (right), the predicted depth map is used to generate the corresponding RQMap.
• Figure 3: Radio signal quality heatmaps in the warehouse with different antenna placements: (a) bottom-left corner, (b) center.
• Figure 4: Evolution of depth maps, predicted RQMaps, and ground-truth RQMaps over time, produced by MapViT.
• Figure 5: (a) Compares Stage 2 inference time of ML-based approaches (ViT, CNN, MLP) against the ray tracer software (RT). (b) Depicts the evaluated regions in Table \ref{['table:regional_loss']}.