Core Concepts
This paper proposes an efficient extreme learning machine (ELM)-based neural network framework to estimate the sensing and communication channels in an IRS-assisted multi-user integrated sensing and communication (ISAC) system.
Abstract
The paper proposes a two-stage channel estimation approach for an IRS-assisted multi-user ISAC system.
In the first stage, the direct sensing and communication channels are estimated at the ISAC base station (BS) and downlink users, respectively, by turning off the IRS. The ISAC BS estimates the sensing channel and uplink communication channels, while each downlink user estimates its direct downlink communication channel.
In the second stage, the reflected uplink and downlink communication channels are estimated by turning on the IRS. The ISAC BS estimates the reflected uplink communication channels, while each downlink user estimates its reflected downlink communication channel.
To realize the proposed two-stage approach, an ELM-based neural network framework is established. Two types of input-output pairs are designed for the ELMs to improve the estimation accuracy and reduce the training complexity. Simulation results demonstrate that the proposed ELM-based approach outperforms the least-squares and neural network-based benchmark schemes in terms of estimation accuracy, training speed, and computational complexity.
Stats
The ISAC BS is equipped with M transmit and M receive antennas.
The IRS has L passive reflecting elements.
There are K uplink users and J downlink users in the system.
Quotes
"The estimation problem in such a system is challenging since the sensing and communication (SAC) signals interfere with each other, and the passive IRS lacks signal processing ability."
"To overcome this problem, a two-stage approach is proposed to transfer the overall estimation problem into sub-ones, successively including the direct and reflected channels estimation."