AI Based Landscape Sensing Using Radio Signals

TL;DR

This paper introduces an AI-based method for landscape sensing using radio signals from a base station, enabling environment inference without the need for radar-equipped user devices, achieving over 95% precision in urban scenarios.

Contribution

It presents a novel AI approach that uses wireless channel features to infer landscapes, eliminating the need for radars on user equipment.

Findings

Achieved over 95% precision in urban landscape sensing.

Demonstrated effectiveness in practical city environments like London.

Provided a cost-effective alternative to radar-based sensing.

Abstract

In many sensing applications, typically radio signals are emitted by a radar and from the bounced reflections of the obstacles, inference about the environment is made. Even though radars can be used to sense the landscapes around the user-equipment (UE) such as whether UE is in the forested region, inside buildings, etc., it is not suitable in many wireless applications as many UEs does not have radars in them. Using radar will also increase the cost and power requirements on the UEs in applications requiring sensing of the landscapes. In this paper, we provide a mechanism where basestation (BS) is able to sense the UE's landscape without the use of a radar. We propose an artificial intelligence (AI) based approach with suitable choice of the features derived from the wireless channel to infer the landscape of the UEs. Results for the proposed methods when applied to practical environments such as London city scenario yields a precision score of more than 95 percent.