KR-20260067970-A - Deep Reinforcement Learning-Based Aerial Reconfigurable Intelligent Surface-Assisted Integrated Sensing and Communication Method and System

KR20260067970AKR 20260067970 AKR20260067970 AKR 20260067970AKR-20260067970-A

Abstract

A deep reinforcement learning-based integrated sensing and communication method and system supporting an air-reconstructed intelligent surface are presented. The deep reinforcement learning-based integrated sensing and communication method supporting an air-reconstructed intelligent surface proposed in the present invention comprises the steps of: collecting location information of multiple users and expected location information of a target to be detected through a collection unit, and collecting information on whether LoS is blocked between a base station, a user, and a target, and channel status information; moving an ARIS along an initial flight path based on the collected information through an ARIS initial deployment unit, and setting an initial transmission beamforming vector and a phase shift of the reconstruction intelligent surface; simultaneously adjusting the flight trajectory of the ARIS, the transmission beamforming vector, and the RIS phase shift for each time slot through a learning unit; suppressing self-interference and clutter echo caused by user reflection occurring while receiving a target reflection signal in full-duplex mode at the base station through a reception signal processing unit, and estimating the coordinates of the target using the suppressed reception signal; and providing communication services to multiple users based on the estimated target coordinates and transmission and reception settings through an integrated sensing and communication providing unit, while simultaneously updating the location of the target.

Inventors

강준혁
유성훈
조현상

Assignees

한국과학기술원

Dates

Publication Date: 20260513
Application Date: 20250811
Priority Date: 20241104

Claims (10)

In an integrated sensing and communication method that detects a target using an Aerial Reconfigurable Intelligent Surface (ARIS) and provides communication services to multiple users in an environment where a Line-of-Sight (LoS) path between a base station and a user is not secured due to obstacles and terrain, The above method is, A step of collecting location information of multiple users and expected location information of a target to be detected through a collection unit, and collecting information on whether LoS is blocked between the base station, the user, and the target, as well as channel status information; A step of moving the ARIS along an initial flight path based on the collected information through the ARIS initial deployment unit, and setting the initial transmit beamforming vector and the phase shift of the Reconfigurable Intelligent Surface (RIS); A step of simultaneously adjusting the flight trajectory, transmission beamforming vector, and RIS phase shift of the ARIS at each time slot through a learning unit; A step of suppressing self-interference and clutter echo caused by user reflection that occur while receiving a target reflection signal in full-duplex mode at the base station through a receiving signal processing unit, and estimating the coordinates of the target using the suppressed receiving signal; and A step of providing communication services to multiple users based on the estimated target coordinates and transmission and reception settings through an integrated sensing and communication provider, while simultaneously updating the target's location. An integrated sensing and communication method that includes aerial reconfiguration intelligent surface support.
In paragraph 1, The step of collecting location information of multiple users and expected location information of a target to be detected through the above-mentioned collection unit, and collecting information on whether LoS is blocked between the base station, the user, and the target, and channel status information, is as follows: Periodically updating the location coordinates of the aforementioned multiple users and targets and collecting channel state information with Rician fading applied, and utilizing it for ARIS flight trajectory and base station beamforming control. Aerial reconstruction intelligent surface support integrated sensing and communication method.
In paragraph 1, The step of simultaneously adjusting the flight trajectory, transmission beamforming vector, and RIS phase shift of the ARIS at each time slot through the learning unit is, Define the flight trajectory, transmitted beamforming vector, and RIS phase shift of the above ARIS as states, and perform learning using the DDPG (Deep Deterministic Policy Gradient) algorithm by setting the Cramer-Rao Bound (CRB), which represents target coordinate estimation accuracy, and multiple user signal-to-interference noise ratios as compensation functions. Aerial reconstruction intelligent surface support integrated sensing and communication method.
In paragraph 1, The step of suppressing self-interference and clutter echo caused by user reflection that occur while receiving target reflection signals in full-duplex mode at the base station through the received signal processing unit, and estimating the coordinates of the target using the suppressed received signal, The above base station performs receiving beamforming using the Null-Space Projection (NSP) method to suppress self-interference signals and clutter echo signals generated from the user terminal, and estimates the coordinates of a target based on the suppressed receiving signals. Aerial reconstruction intelligent surface support integrated sensing and communication method.
In paragraph 1, The step of providing communication services to multiple users based on the estimated target coordinates and transmission and reception settings through the integrated sensing and communication provider, while simultaneously updating the target's location, is: The method dynamically updates the flight trajectory of the ARIS based on the target coordinate estimation results, and repeatedly performs the step of simultaneously adjusting the flight trajectory of the ARIS, the transmit beamforming vector, and the RIS phase shift for each time slot through the learning unit, and simultaneously improves the detection accuracy and communication transmission rate in the next time slot by resetting the transmit beamforming and RIS phase shift based on the updated flight trajectory of the ARIS. Aerial reconstruction intelligent surface support integrated sensing and communication method.
In an integrated sensing and communication system that detects targets using an Aerial Reconfigurable Intelligent Surface (ARIS) and provides communication services to multiple users in an environment where a Line-of-Sight (LoS) path between a transmitting and receiving base station and a user is not secured due to obstacles and terrain, A collection unit that collects location information of multiple users and expected location information of a target to be detected, and collects information on whether LoS is blocked between the base station, the user, and the target, as well as channel status information; An ARIS initial deployment unit that moves the ARIS along an initial flight path based on the collected information and sets the initial transmit beamforming vector and the phase shift of the Reconfigurable Intelligent Surface (RIS); A learning unit that simultaneously adjusts the flight trajectory, transmission beamforming vector, and RIS phase shift of the above ARIS for every time slot; A receiving signal processing unit that suppresses self-interference and clutter echo caused by user reflection occurring while receiving target reflection signals in full-duplex mode at the base station, and estimates the coordinates of the target using the suppressed receiving signal; and An integrated sensing and communication providing unit that provides communication services to multiple users and simultaneously updates the target's location based on the above-mentioned estimated target coordinates and transmission and reception settings. An aerial reconfigurable intelligent surface-assisted integrated sensing and communication system including
In paragraph 6, The above learning unit is, Define the flight trajectory, transmitted beamforming vector, and RIS phase shift of the above ARIS as states, and perform learning using the DDPG (Deep Deterministic Policy Gradient) algorithm by setting the Cramer-Rao Bound (CRB), which represents target coordinate estimation accuracy, and multiple user signal-to-interference noise ratios as compensation functions. Aerial reconfigurable intelligent surface support integrated sensing and communication system.
In paragraph 6, The above-mentioned received signal processing unit is, The above base station performs receiving beamforming using the Null-Space Projection (NSP) method to suppress self-interference signals and clutter echo signals generated from the user terminal, and estimates the coordinates of a target based on the suppressed receiving signals. Aerial reconfigurable intelligent surface support integrated sensing and communication system.
In paragraph 6, The above integrated sensing and communication providing unit is, The method dynamically updates the flight trajectory of the ARIS based on the target coordinate estimation results, and repeatedly performs the step of simultaneously adjusting the flight trajectory of the ARIS, the transmit beamforming vector, and the RIS phase shift for each time slot through the learning unit, and simultaneously improves the detection accuracy and communication transmission rate in the next time slot by resetting the transmit beamforming and RIS phase shift based on the updated flight trajectory of the ARIS. Aerial reconfigurable intelligent surface support integrated sensing and communication system.
A program stored on a computer-readable storage medium for executing a deep reinforcement learning-based aerial reconstruction intelligent surface support integrated sensing and communication method, A step of collecting location information of multiple users and expected location information of a target to be detected through a collection unit, and collecting information on whether LoS is blocked between the base station, the user, and the target, as well as channel status information; A step of moving the ARIS along an initial flight path based on the collected information through the ARIS initial deployment unit, and setting the initial transmit beamforming vector and the phase shift of the Reconfigurable Intelligent Surface (RIS); A step of simultaneously adjusting the flight trajectory, transmission beamforming vector, and RIS phase shift of the ARIS at each time slot through a learning unit; A step of suppressing self-interference and clutter echo caused by user reflection that occur while receiving a target reflection signal in full-duplex mode at the base station through a receiving signal processing unit, and estimating the coordinates of the target using the suppressed receiving signal; and A step of providing communication services to multiple users based on the estimated target coordinates and transmission and reception settings through an integrated sensing and communication provider, while simultaneously updating the target's location. A program stored on a computer-readable storage medium containing

Description

Deep Reinforcement Learning-Based Aerial Reconfigurable Intelligent Surface-Assisted Integrated Sensing and Communication Method and System The present invention relates to a deep reinforcement learning-based aerial reconstruction intelligent surface support integrated sensing and communication method and system. As we transition beyond 5G and 6G, the increasing number of users is driving a growing demand for the efficient utilization of frequencies. In particular, Integrated Sensing and Communication (ISAC) for unmanned aerial vehicles is gaining prominence as a promising technology for future wireless networks that utilizes high-frequency bands such as millimeter wave (mmWave). Given that radar sensing and wireless communication share the same spectrum and hardware, ISAC can improve communication and sensing efficiency in environments with poor infrastructure conditions. The overall process of an ISAC downlink system typically involves a transceiver transmitting ISAC signals to a user and processing echo signals reflected from the target to be detected. However, in the downlink, the Line of Sight (LoS) is blocked by obstacles such as mountains and buildings, making it impossible to avoid problems such as severe path loss and attenuation as communication distances increase. To overcome the physical limitations of the Line of Sight (LoS), research is being developed on Reconfigurable Intelligent Surfaces (RIS), a core technology that extends target detection and communication range by reconfiguring signal propagation and adjusting phase shifts. However, ground-based RISs, which are nodes between transceivers and users, have limitations in supporting multiple users and in more complex detection and communication environments. On the other hand, an Aerial Reconfigurable Intelligent Surface (ARIS), which is an unmanned aerial vehicle (UAV) equipped with an RIS, can provide effective detection and communication performance in more dynamic and poor infrastructure environments by leveraging the mobility of the UAV. Solutions in related research dealing with ARIS systems for sensing or communication networks are mostly provided by convex optimization. However, in situations involving multiple users, IoT services targeting beyond 5G/6G handle applications that require high computational power. Furthermore, due to time-varying factors such as channel conditions, existing convex optimization-based solutions face the problem of high complexity in optimizing ARIS trajectories. Although deep learning-based reinforcement learning methods have been proposed as a representative countermeasure, previous related studies have failed to solve the problem of continuous control. Therefore, there is a need to propose a new optimization method that can be well applied to environments that continuously change over time while securing higher sensing and communication transmission performance. Additionally, to implement a practical ISAC, a method is required to suppress self-interference signals that occur when transceivers operate in full-duplex mode, as well as clutter echoes, which are reflection signals generated from clutter rather than the target to be recognized. FIG. 1 is a diagram illustrating a system environment for performing integrated sensing and communication through the trajectory movement of an aerial reconfigurable intelligent surface ARIS according to one embodiment of the present invention. FIG. 2 is a diagram showing the configuration of a deep reinforcement learning-based aerial reconstruction intelligent surface support integrated sensing and communication system according to one embodiment of the present invention. FIG. 3 is a flowchart illustrating a deep reinforcement learning-based aerial reconstruction intelligent surface support integrated sensing and communication method according to one embodiment of the present invention. FIG. 4 is a diagram showing the accumulated reward value when reinforcement learning according to one embodiment of the present invention is applied. FIG. 5 is a diagram showing the squared estimation error for a target position received by a transceiver according to an embodiment of the present invention over a time frame. FIG. 6 is a diagram showing the ARIS path and the estimated location of the target for each time frame according to an embodiment of the present invention. The present invention relates to an integrated sensing and communication system using an ARIS, in which a multi-antenna transceiver base provides communication services to ground users and detects targets. To ensure efficient integrated sensing and communication performance, a method is proposed to jointly optimize the ARIS trajectory, transmit beamforming, and RIS phase shift while simultaneously guaranteeing the communication transmission rate of users. To address this, a deep learning-based reinforcement learning method is applied, and it is demonstrated that optimal performance is achieved at every time fram