Two-Stage Radio Map Construction with Real Environments and Sparse Measurements

Yifan Wang,Shu Sun,Na Liu,Lianming Xu,Li Wang

2024-10-08

Abstract:Radio map construction based on extensive measurements is accurate but expensive and time-consuming, while environment-aware radio map estimation reduces the costs at the expense of low accuracy. Considering accuracy and costs, a first-predict-then-correct (FPTC) method is proposed by leveraging generative adversarial networks (GANs). A primary radio map is first predicted by a radio map prediction GAN (RMP-GAN) taking environmental information as input. Then, the prediction result is corrected by a radio map correction GAN (RMC-GAN) with sparse measurements as guidelines. Specifically, the self-attention mechanism and residual-connection blocks are introduced to RMP-GAN and RMC-GAN to improve the accuracy, respectively. Experimental results validate that the proposed FPTC-GANs method achieves the best radio map construction performance, compared with the state-of-the-art methods.

Signal Processing,Artificial Intelligence

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to reduce cost and time while ensuring accuracy when constructing radio maps. Although the traditional method based on a large number of measurements is accurate, it is very expensive and time - consuming. And the radio map estimation method based on environmental awareness reduces the cost but sacrifices accuracy. For this reason, the paper proposes a new two - stage method - First - Predict - Then - Correct (FPTC), which uses Generative Adversarial Networks (GANs) to improve the construction performance of radio maps.

Specifically, this method first predicts the initial radio map according to environmental information through a Radio Map Prediction GAN (RMP - GAN), and then corrects the prediction result by using sparse measurement data through another Radio Map Correction GAN (RMC - GAN) to reflect the real - time communication quality. In order to improve the accuracy of the model, the RMP - GAN introduces the Self - Attention Mechanism, and the RMC - GAN adopts Residual - Connection Blocks.

The experimental results show that the proposed FPTC - GANs method is superior to the existing state - of - the - art methods in the performance of constructing radio maps, achieving the best balance between cost and accuracy.