# US 2026/0074748 A1 · Method for joint active and passive beamforming and received signal optimization in isac system assisted by dual irss

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0074748 A1
- Document ID: US20260074748A1
- Status: Published application
- Kind: A1
- Application number: 19/389,095
- Filed: 2025-11-14
- Published: 2026-03-12
- Earliest claimed: 2024-05-24
- Assignees: Nanjing University of Posts and Telecommunications
- Applicants: Nanjing University of Posts and Telecommunications
- Inventors: Chengyin XIONG; Fei LI; Yuanyuan WANG; Ting LI
- Web page: https://patentgrep.com/patent/US20260074748A1

## CPC

- H04B 7/04013: {Intelligent reflective surfaces}
- H04B 7/0617: {for beam forming}
- H04B 17/346: {Noise values (signal-to-interference ratio [SIR] or carrier-to-interference ratio [CIR] H04B17/336)}
- H04B: TRANSMISSION

## Abstract

A method for the joint active and passive beamforming and received signal optimization in an ISAC system assisted by dual IRSs is provided. The method jointly optimizes active beamforming at Base Station (BS), reception of sensing signals at the Base Station (BS), and passive beamforming at IRSs, so as to maximize communication sum-rate of users while ensuring that SNR of sensing signals meets a minimum requirement. To address the complex non-convex optimization problem, the method first applies fractional programming to decouple problem, then adopts successive convex approximation algorithm and alternating direction method of multipliers to transform intractable non-convex problem into multiple tractable subproblems, and finally employs an alternating optimization method to efficiently acquire the high-quality suboptimal solutions. The simulation results demonstrate that disclosed scheme exhibits satisfactory convergence and effectiveness, and can significantly improve the performance of IRS-assisted ISAC systems.

## Claims

### Claim 1

A method for joint active and passive beamforming and received signal optimization in an integrated sensing and communication (ISAC) system assisted by dual Intelligent Reflecting Surfaces (IRSs), comprising:
S1, model establishing:
establishing an ISAC system model assisted by the dual IRSs, wherein the ISAC system model comprises a multi-antenna Base Station (BS) configured to simultaneously provide communication services for a plurality of users and perform sensing for targets, wherein the dual IRSs assist the multi-antenna BS in operation, and reliable channels between the multi-antenna BS and communication users and sensing targets are established by dispersing the dual IRSs on a surface of a building; and
S2, alternating optimization:
S21: deriving a signal-to-interference-plus-noise ratio (SINR) for the communication users and a signal-to-noise ratio (SNR) lower bound for the sensing targets based on the reliable channels for communication and sensing, and formulating an optimization problem based on the SINR and the SNR lower bound, the optimization problem jointly designing active beamforming w at a BS end, received signal processing u at the multi-antenna BS for a sensing signal, and passive beamforming θ1,θ2 at the dual IRSs, under an objective of maximizing a communication sum-rate of the communication users, while ensuring that a sensing SNR satisfies a minimum requirement, and satisfying a power constraint at the BS end and a constant modulus constraint at IRSs;
S22: introducing auxiliary variables β and ρ, and decoupling the optimization problem using a fractional programming method;
S23: updating the auxiliary variables β and ρ by obtaining partial derivatives, and updating the received signal processing u using Rayleigh quotient under given premise conditions w,θ1,θ2,ρ,β;
S24: representing an original optimization problem as a second-order cone programming (SOCP) problem under given conditions u,ρ,β,θ1,θ2, and solving and updating the active beamforming w using a convex optimization (CVX) toolbox;
S25: introducing an auxiliary variable φ as well as a dual variable μ, successively updating first reflection coefficient vectors θ1,φ, μ using an alternating direction method of multipliers (ADMM) under given conditions u,ρ,β,w,θ2, and updating a second reflection coefficient vector θ2 using a same method;
S26: repeating S25 until values of θ1,θ2 converges; and
S27: repeating S23, S24, S25, and S26 until an algorithm converges, thereby obtaining an optimal solution wopt,uopt,
θ
1
opt
,
θ
2
opt
.

### Claim 2 (depends on 1)

The method for the joint active and passive beamforming and the received signal optimization in the ISAC system assisted by the dual IRSs according to claim 1, wherein the ISAC system model assisted by the dual IRSs in the S1 is defined as:
in the ISAC system model assisted by the dual IRSs, a multi-antenna BS simultaneously performs multi-user communication and target detection with assistance of two N-element IRSs, specifically:
a multi-antenna BS end is equipped with M transmit antennas as well as M receive antennas, arranged in a uniform linear array (ULA) with a half-wavelength interval, and simultaneously transmits data to K single-antenna users while detecting T sensing targets, wherein a transmit signal of the multi-antenna Base Station (BS) is represented as:
x
=
W
⁢
s
=
W
c
⁢
s
c
+
W
r
⁢
s
r
wherein We∈M×K as well as Wr∈M×M are, respectively, a communication beamforming matrix and a sensing beamforming matrix, sc∈K is a vector representing communication signals satisfying
E
⁢
{
s
c
⁢
s
c
H
}
=
I
K
,
sr∈M is a vector representing sensing signals satisfying
E
⁢
{
s
r
⁢
s
r
H
}
=
I
M
,
assuming they are statistically independent of each other and satisfying
E
⁢
{
s
c
⁢
s
r
H
}
=
0
,
W[Wc Wr]∈M×(K+M) is defined as an entire beamforming matrix, and
s
⁢
[
s
c
T
⁢
s
r
T
]
T
∈
K
+
M
is defined as a transmit symbol vector.

### Claim 3 (depends on 1)

The method for the joint active and passive beamforming and the received signal optimization in the ISAC system assisted by the dual IRSs according to claim 1, wherein the reliable channels in the S1 specifically comprise:
considering the ISAC system model assisted by the dual IRSs is in a crowded environment, a channel between a BS and the IRSs is modeled as a Rician fading channel, represented as:
G
=
ω
R
1
+
ω
R
⁢
G
LoS
)
+
1
1
+
ω
R
⁢
G
NLoS
,
wherein ωR is a Rician factor of a BS-IRS link, which degenerates to a Line-of-Sight (LoS) scenario when ωR approaches infinity, and becomes a Rayleigh channel when ωR approaches zero, GNLoS∈N×M is a Rayleigh fading component, each of a term satisfying CN(0,1) distribution, GLos∈N×M is an LoS channel component, because the multi-antenna BS and the IRSs are modeled as, respectively, a uniform linear array (ULA) and a uniform planar array (UPA), and a channel matrix GLoS is represented as:
G
L
⁢
o
⁢
S
=
α
⁢
e
j
⁢
θ
⁢
a
R
(
θ
R
)
⁢
a
T
H
(
θ
T
)
,
wherein α is a large-scale channel gain, θ is a random phase uniformly distributed in [0,2π], aT is a transmit steering vector of the BS, and aR is a receive steering vector of the IRSs, wherein a steering vector is represented as a(θ)=[e−j2πd sin(θ)/λ, . . . , e−j2πd(M−1)sin(θ)/λ]T wherein d is a distance between array elements, which is typically set to a half-wavelength, and λ is a signal wavelength;
a channel between the multi-antenna BS and each user is composed of two parts, which are a direct link between the BS and a user and a cascaded link between the BS, the IRSs, and the user, therefore, a received signal at a kth user from the multi-antenna BS is represented as:
y
k
=
(
H
k
H
+
F
1
,
k
H
⁢
Θ
1
H
⁢
G
1
+
F
2
,
k
H
⁢
Θ
2
H
⁢
G
2
)
⁢
x
+
n
k
,
wherein Hk∈M, F1,k∈N, F2,k∈N, G1∈N×M, G2∈N×M are defined as effective channels between the BS and the kth user, between IRS1 and the kth user, between IRS2 and the kth user, between the BS and the IRS1, and between the BS and the IRS2, respectively, and a reflection matrix of an IRS is defined as Θ1diag{θi}, wherein θi=[θi1, . . . , θiN]T is a reflection coefficient vector satisfying |θin|=1, ∀i∈{1,2}, ∀n.

### Claim 4 (depends on 1)

The method for the joint active and passive beamforming and the received signal optimization in the ISAC system assisted by the dual IRSs according to claim 1, wherein the SINR in the S21 is defined as:
a scalar
n
k
CN
⁡
(
0
,
σ
k
2
)
is an additive white gaussian noise (AWGN) at a kth user, therefore, the SINR of the kth user is calculated as:
γ
k
=
❘
"\[LeftBracketingBar]"
h
k
H
⁢
w
k
❘
"\[RightBracketingBar]"
2
∑
j
≠
k
K
+
M
⁢
❘
"\[LeftBracketingBar]"
h
k
H
⁢
w
j
❘
"\[RightBracketingBar]"
2
+
σ
k
2
,
wherein an effective channel hk∈M is defined as
h
k
H
=
H
k
H
+
F
1
,
k
H
⁢
Θ
1
H
⁢
G
1
+
F
2
,
k
H
⁢
Θ
2
H
⁢
G
2
,
wherein wj denotes a jth column of a beamforming matrix W, i.e., W[w1, . . . , wK+M];
considering a link between the multi-antenna BS and a th target is blocked, an echo signal received through a path assisted by the IRSs is represented as:
y
r
,
t
=
α
t
(
G
1
H
⁢
Θ
1
⁢
z
1
,
t
+
G
2
H
⁢
Θ
2
⁢
z
2
,
t
)
⁢
(
z
1
,
t
H
⁢
Θ
1
H
⁢
G
1
+
z
2
,
t
H
⁢
Θ
2
H
⁢
G
2
)
⁢
W
⁢
s
+
n
r
,
further considering target detection is performed under far-field conditions, wherein αt represents radar cross section (RCS) of the tth target, i.e., αt=4πP⊏/S, in which P⊏ denotes a radiated power density of a target's scattered wave, and S denotes a power density of an incident wave; calculating a sensing target SNR in terms of expectation; for simplifying a formula expression, utilizing
σ
t
2
to represent an expected value αt, i.e.,
E
⁢
{
❘
"\[LeftBracketingBar]"
α
t
❘
"\[RightBracketingBar]"
2
}
=
σ
t
2
,
wherein z1,t∈N and z2,t∈N denote baseband channels between IRS1, IRS2 and the tth target, respectively, and a vector
n
r
⁢
□
⁢
C
⁢
N
⁡
(
0
,
σ
r
2
⁢
I
M
)
represents the AWGN; assuming the path between the IRSs and a target is LoS, and required angles of arrival/departure (AoA/AoD) are known, an equivalent channel matrix of the echo signal H, is redefined as:
H
t
(
G
1
H
⁢
Θ
1
⁢
z
1
,
t
+
G
2
H
⁢
Θ
2
⁢
z
2
,
t
)
⁢
(
z
1
,
t
H
⁢
Θ
1
H
⁢
G
1
+
z
2
,
t
H
⁢
Θ
2
H
⁢
G
2
)
;
defining wvec{W} as vectorizing a matrix W, and ⊗ as a Kronecker product, then yr,t is re-expressed as:
y
r
,
t
=
α
t
(
S
⁢
S
H
⊗
H
t
)
⁢
w
+
n
r
.

### Claim 5 (depends on 1)

The method for the joint active and passive beamforming and the received signal optimization in the ISAC system assisted by the dual IRSs according to claim 1, wherein the SNR lower bound in the S21 is defined as:
configuring the multi-antenna BS to process a received sensing signal yr,t as:
u
H
⁢
y
r
,
t
=
α
t
⁢
u
H
(
S
⁢
S
H
⊗
H
t
)
⁢
w
+
u
H
⁢
n
r
,
therefore, an SNR of a tth sensing target is calculated as:
γ
r
,
t
=
σ
t
2
⁢
E
⁢
{
❘
"\[LeftBracketingBar]"
u
H
(
S
⁢
S
H
⊗
H
t
)
⁢
w
❘
"\[RightBracketingBar]"
2
}
σ
r
2
⁢
u
H
⁢
u
;
for simplifying a formula expression, redefining E{SSH}=IK+M; utilizing a Jensen's inequality, i.e., E{ƒ(x)}≥ƒ(E{x}), a resulting SNR lower bound is obtained as:
r
r
,
t
≥
σ
t
2
⁢
❘
"\[LeftBracketingBar]"
u
H
(
I
K
+
M
⊗
H
t
)
⁢
w
❘
"\[RightBracketingBar]"
2
σ
r
2
⁢
u
H
⁢
u
.

### Claim 6 (depends on 1)

The method for the joint active and passive beamforming and the received signal optimization in the ISAC system assisted by the dual IRSs according to claim 1, wherein the optimization problem in the S21 is defined as: multi-antenna BS received signal processing configuration u and a dual IRSs passive beamforming Θ1 as well as Θ2 are optimized to maximize the communication sum-rate of a multi-user, while simultaneously satisfying a worst-case sensing SNR Γt, a transmit power budget P and a unit modulus constraint of reflection coefficients; therefore, the optimization problem is formulated as:
max
Θ
,
W
R
s
⁢
u
⁢
m
(
Θ
,
W
)
=
∑
k
=
1
K
log
2
(
1
+
γ
k
)
s
.
t
.
C
1
:
γ
r
,
t
≥
Γ
t
,
∀
t
C
2
:

W

F
2
≤
P
C
3
:
❘
"\[LeftBracketingBar]"
θ
i
,
n
❘
"\[RightBracketingBar]"
=
1
,
∀
i
∈
{
1
,
2
}
,
∀
n
∈
N
.

### Claim 7 (depends on 6)

The method for the joint active and passive beamforming and the received signal optimization in the ISAC system assisted by the dual IRSs according to claim 6, wherein the optimization problem is converted and optimized through the S22 to the S24, specifically comprising:
firstly, introducing the auxiliary variable ρ=[ρ1, ρ2, . . . , ρK]T via Lagrangian duality transformation, and converting an objective function to the following form:
∑
k
=
1
K
log
2
(
1
+
ρ
k
)
-
∑
k
=
1
K
ρ
k
+
∑
k
=
1
K
(
1
+
ρ
k
)
⁢
❘
"\[LeftBracketingBar]"
h
k
H
⁢
v
k
❘
"\[RightBracketingBar]"
2
∑
j
=
1
K
+
M
⁢
❘
"\[LeftBracketingBar]"
h
k
H
⁢
w
j
❘
"\[RightBracketingBar]"
2
+
σ
k
2
,
by introducing the auxiliary variable β=[β1, β2, . . . , βK]T, expanding the objective function to a quadratic form:
f
⁡
(
w
,
θ
,
ρ
,
β
)
=
∑
k
=
1
K
log
2
(
1
+
ρ
k
)
-
∑
k
=
1
K
ρ
k
-
∑
k
=
1
K
❘
"\[LeftBracketingBar]"
β
k
❘
"\[RightBracketingBar]"
2
⁢
σ
k
2
+
∑
k
=
1
K
2
⁢
1
+
ρ
k
⁢
ℜ
⁢
{
β
k
*
⁢
h
k
T
⁢
w
k
}
-
∑
k
=
1
K
❘
"\[LeftBracketingBar]"
β
k
❘
"\[RightBracketingBar]"
2
⁢
∑
j
=
1
K
+
M
❘
"\[LeftBracketingBar]"
h
k
H
⁢
w
j
❘
"\[RightBracketingBar]"
2
,
then a new objective function becomes ƒ(w,θ,ρ,β), which is converted into a more concise form through equivalent transformation:
f
⁡
(
w
,
θ
,
ρ
,
β
)
=
ℜ
⁢
{
v
H
⁢
w
}
-

Bw

2
+
ε
1
=
ℜ
⁢
{
g
H
⁢
θ
}
-
θ
H
⁢
Λθ
+
ε
2
;
in an above-mentioned formula,
w
⁢
□
⁢
v
⁢
e
⁢
c
⁢
{
W
}
=
[
w
1
T
,
w
2
T
,
…
,
w
K
+
M
T
]
T
is defined, wherein w∈M(K+M)×1 and wj∈M×1, and extraction of wj from w is achieved by defining a permutation matrix Qj∈M×M(K+M); θ is utilized to represent θ1 or θ2 for simplifying an expression; an equivalent expression of ƒ(w,θ,ρ,β) is obtained by applying the formula
h
k
H
⁢
w
j
=
H
k
H
⁢
w
j
+
F
1
,
k
H
⁢
diag
⁢
{
G
1
⁢
w
j
}
⁢
θ
1
H
+
F
2
,
k
H
⁢
diag
⁢
{
G
2
⁢
w
j
}
⁢
θ
2
H
;
remaining variables other than w are defined as follows:
v
=
[
2
⁢
1
+
ρ
k
⁢
β
k
H
⁢
h
k
H
,
…
,
2
⁢
1
+
ρ
K
⁢
β
K
H
⁢
h
K
H
,
0
H
]
H
B
=
[
b
1
,
1
,
b
1
,
2
,
…
,
b
1
,
K
+
M
,
…
,
b
K
,
1
,
b
K
,
2
,
…
,
b
K
,
K
+
M
]
T
,
b
k
,
j
⁢
□
⁢
❘
"\[LeftBracketingBar]"
β
k
❘
"\[RightBracketingBar]"
⁢
Q
j
H
⁢
h
k
ε
1
=
∑
k
=
1
K
log
2
(
1
+
ρ
k
)
-
∑
k
=
1
K
ρ
k
-
∑
k
=
1
K
❘
"\[LeftBracketingBar]"
β
k
❘
"\[RightBracketingBar]"
2
⁢
σ
k
2
;
similarly, remaining variables other than θ are defined as:
g
2
⁢
∑
k
=
1
K
1
+
ρ
k
⁢
β
k
(
diag
⁢
{
w
k
H
⁢
G
1
H
}
⁢
F
1
,
k
+
diag
⁢
{
w
k
H
⁢
G
2
H
}
⁢
F
2
,
k
)
-
2
⁢
∑
k
=
1
K
❘
"\[LeftBracketingBar]"
β
k
❘
"\[RightBracketingBar]"
2
⁢
∑
j
=
1
K
+
M
(
diag
⁢
{
w
j
H
⁢
G
1
H
}
⁢
F
1
,
k
⁢
H
k
H
⁢
w
j
+
diag
⁢
{
w
j
H
⁢
G
2
H
}
⁢
F
2
,
k
⁢
H
k
H
⁢
w
j
)
A
∑
k
=
1
K
❘
"\[LeftBracketingBar]"
β
k
❘
"\[RightBracketingBar]"
2
⁢
∑
j
=
1
K
+
M
(
diag
⁢
{
w
j
H
⁢
G
1
H
}
⁢
F
1
,
k
⁢
F
1
,
k
H
⁢
diag
⁢
{
G
1
⁢
w
j
}
+
diag
⁢
{
w
j
H
⁢
G
2
H
}
⁢
F
2
,
k
⁢
F
2
,
k
H
⁢
diag
⁢
{
G
2
⁢
w
j
}
)
ε
2
ε
1
+
∑
k
=
1
K
[
2
⁢
1
+
ρ
k
⁢
ℜ
⁢
{
β
k
H
⁢
H
k
H
⁢
w
k
}
-
❘
"\[LeftBracketingBar]"
β
k
❘
"\[RightBracketingBar]"
2
⁢
∑
j
=
1
K
+
M
❘
"\[LeftBracketingBar]"
H
k
H
⁢
w
j
❘
"\[RightBracketingBar]"
2
]
.

### Claim 8 (depends on 6)

The method for the joint active and passive beamforming and the received signal optimization in the ISAC system assisted by the dual IRSs according to claim 6, wherein an alternating optimization method is adopted in the S24 to the S27 in order to iteratively solve respective optimization variables, specifically comprising:
before optimizing a configuration u of the multi-antenna BS for received signals, both the auxiliary variables ρ as well as β are optimized first, and under given conditions β, u, w, θ1 and θ2, optimization of the auxiliary variable ρ is an unconstrained convex problem, and by calculating a partial derivative ∂ƒ/∂ρk=0, an optimal solution for an auxiliary variable
ρ
k
opt
is obtained as:
ρ
k
opt
=
γ
k
*
=
❘
"\[LeftBracketingBar]"
h
k
H
⁢
w
k
❘
"\[RightBracketingBar]"
2
∑
j
≠
k
K
+
M
⁢
❘
"\[LeftBracketingBar]"
h
k
H
⁢
w
j
❘
"\[RightBracketingBar]"
2
+
σ
k
2
,
∀
k
;
similarly, given
ρ
k
opt
,
u, w, θ1 and θ2, by setting ∂ƒ/∂βk=0, an optimal solution
β
k
opt
is obtained as:
β
k
opt
=
1
+
ρ
k
opt
⁢
(
∑
j
=
1
K
+
M
❘
"\[LeftBracketingBar]"
h
k
H
⁢
w
j
❘
"\[RightBracketingBar]"
2
+
σ
k
2
)
-
1
⁢
h
k
H
⁢
w
k
,
∀
k
;
next, fixing other variables, a multi-antenna BS configuration u is then optimized specifically as follows:
firstly, a maximization problem is defined as:
max
u
σ
t
2
⁢
❘
"\[LeftBracketingBar]"
u
H
(
I
K
+
M
⊗
H
t
)
⁢
w
❘
"\[RightBracketingBar]"
2
σ
r
2
⁢
u
H
⁢
u
,
∀
t
,
and an optimal solution uopt is derived through knowledge of the Rayleigh quotient, and a result is:
u
opt
=
(
I
K
+
M
⊗
H
t
)
⁢
w
w
H
(
I
K
+
M
⊗
H
t
H
⁢
H
t
)
⁢
w
,
∀
t
;
under the given conditions ρ, β, u, θ1 as well as θ2, optimization of transmit beamforming w is represented as:
min
w

Bw

2
-
{
v
H
⁢
w
}
s
.
t
.
C
1
:
γ
r
,
t
≥
Γ
t
,
∀
t
C
2
:

w

2
≤
P
,
a sensing constraint C1 regarding an objective function is a non-convex function, in order to handle this non-convex constraint, C1 is represented as a form of second-order cone constraint (SOCP), which is formulated as:
{
u
H
(
I
K
+
M
⊗
H
t
)
⁢
w
}
≥
σ
r
2
⁢
u
H
⁢
u
⁢
Γ
t
/
σ
t
2
,
the optimization problem is redefined as:
min
w

Bw

2
-
{
v
H
⁢
w
}
s
.
t
.
C
1
:
{
u
H
(
I
K
+
M
⊗
H
t
)
⁢
w
}
≥
σ
r
2
⁢
u
H
⁢
u
⁢
Γ
t
/
σ
r
2
,
∀
t
,
C
2
:

w

2
≤
P
then the optimization problem is redefined as a simple convex problem, which can be solved by the CVX toolbox; under the given conditions ρ, β, u, w and θ2, an optimization problem regarding a reflection coefficient θ1 is expressed as:
min
θ
1
θ
1
H
⁢
Λθ
1
-
{
g
H
⁢
θ
1
}
s
.
t
.
C
1
:
{
u
H
(
I
K
+
M
⊗
H
t
)
⁢
w
}
≥
σ
r
2
⁢
u
H
⁢
u
⁢
Γ
t
/
σ
t
2
,
∀
t
;
C
2
:
❘
"\[LeftBracketingBar]"
θ
1
⁢
n
❘
"\[RightBracketingBar]"
=
1
,
∀
n
because both a radar constraint and a non-convex unit modulus constraint involve implicit functions of θ1, which cannot be solved directly, firstly, the non-convex constraint C1 is handled by rewriting an expression on a left-hand side of C1 concerning θ1, and then finding a surrogate function of C1 through a successive convex approximation (SCA) method, specifically comprising:
firstly, an expression Ht is expanded as:
H
t
=
G
1
H
⁢
Θ
1
⁢
z
1
,
t
⁢
z
1
,
t
H
⁢
Θ
1
H
⁢
G
1
+
G
1
H
⁢
Θ
1
⁢
z
1
,
t
⁢
z
2
,
t
H
⁢
Θ
2
H
⁢
G
2
+
G
2
H
⁢
Θ
2
⁢
z
2
,
t
⁢
z
1
,
t
H
⁢
Θ
1
H
⁢
G
1
+
G
2
H
⁢
Θ
2
⁢
z
2
,
t
⁢
z
2
,
t
H
⁢
Θ
2
H
⁢
G
2
,
by utilizing equivalent transformation Θ1z1,t=diag{z1,t}θ1, Θ1z2,t=diag{z2,t}θ1, Θ2z2,t=diag{z2,t}θ2, Θ2z1,t=diag{z1,t}θ2 and vectorized sandwich formula extraction vec(ABC)=(CT⊗A)vec{B}, (IK+M⊗Ht)w is reformulated as:
(
I
K
+
M
⊗
H
t
)
⁢
w
=
∑
i
=
1
2
∑
j
=
1
2
(
I
K
+
M
⁢
G
i
H
⁢
diag
⁢
{
z
i
,
t
}
⁢
θ
i
⁢
θ
j
H
⁢
diag
⁢
{
z
j
,
t
H
}
⁢
G
j
)
⁢
w
=
∑
i
=
1
2
∑
j
=
1
2
vec
⁢
{
G
i
H
⁢
diag
⁢
{
z
i
,
t
}
⁢
θ
i
⁢
θ
j
H
⁢
diag
⁢
{
z
j
,
t
H
}
⁢
G
j
⁢
W
}
=
∑
i
=
1
2
∑
j
=
1
2
(
W
T
⁢
G
j
T
⁢
diag
⁢
{
z
j
,
t
*
}
⊗
G
i
H
⁢
diag
⁢
{
z
i
,
t
}
)
⁢
vec
⁢
{
θ
i
⁢
θ
j
H
}
;
to simplify the expression, an expression
L
1
,
t
=
W
T
⁢
G
1
T
⁢
diag
⁢
{
z
1
,
t
*
}
⊗
G
1
H
⁢
diag
⁢
{
z
1
,
t
}
⁢
and
E
1
,
t
=
W
T
⁢
G
2
T
⁢
diag
⁢
{
z
2
,
t
*
}
⊗
G
1
H
⁢
diag
⁢
{
z
1
,
t
}
+
W
T
⁢
G
1
T
⁢
diag
⁢
{
z
1
,
t
*
}
⊗
G
2
H
⁢
diag
⁢
{
z
2
,
t
}
are defined, given θ2 is fixed, and E1,t∈M(K+M)×N, a constraint then turns into an expression as
⁠
{
u
H
⁢
E
1
,
t
⁢
θ
1
+
u
H
⁢
L
1
,
t
⁢
vec
⁢
{
θ
1
⁢
θ
1
H
}
=
{
u
H
⁢
E
1
,
t
⁢
θ
1
+
θ
1
H
⁢
L
1
,
t
⁢
θ
1
}
≥
σ
r
2
⁢
u
H
⁢
u
⁢
Γ
t
/
σ
t
2
,
wherein
L
1
,
t
=
(
u
H
⁢
L
1
,
t
)
H
=
L
1
,
t
H
⁢
u
,
L1,t∈N×N, since θ1HL1,tθ1 is a convex function of θ1, a lower bound is represented by utilizing an SCA algorithm:
θ
1
H
⁢
L
1
,
t
⁢
θ
1
≥
-
θ
1
(
n
)
⁢
H
⁢
L
1
,
t
⁢
θ
1
(
n
)
+
2
{
θ
1
H
⁢
L
1
,
t
⁢
θ
1
(
n
)
}
,
next,
q
t
⁢
▯
⁢
2
⁢
L
1
,
t
⁢
θ
1
(
n
)
+
E
1
,
t
H
⁢
u
is redefined, then the sensing constraint C1 is reformulated as:
{
θ
1
H
⁢
q
t
}
≥
δ
r
2
⁢
u
H
⁢
u
⁢
Γ
t
/
σ
t
2
+
θ
1
(
n
)
⁢
H
⁢
L
1
,
t
⁢
θ
1
(
n
)
=
ε
3
;
hereafter, an ADMM algorithm is utilized to solve a constant modulus constraint problem, specifically comprising:
firstly, an auxiliary variable φ[φ1, φ2, . . . , φN]T is introduced to convert an optimization problem to be solved θ1 into:
min
θ
1
f
2
(
θ
1
)
=
θ
1
H
⁢
Λ
⁢
θ
1
-
{
g
H
⁢
θ
1
}
s
.
t
.
C
1
:
{
θ
1
H
⁢
q
t
}
≥
ε
3
,
∀
t
C
2
:
❘
"\[LeftBracketingBar]"
θ
1
,
n
❘
"\[RightBracketingBar]"
≤
1
,
∀
n
∈
N
C
3
:
θ
=
φ
C
4
:
❘
"\[LeftBracketingBar]"
φ
n
❘
"\[RightBracketingBar]"
≤
1
,
∀
n
∈
N
;
by utilizing the ADMM algorithm, the problem is further converted through an augmented Lagrangian function as:
min
θ
,
φ
,
μ
θ
1
H
⁢
Λ
⁢
θ
1
-
{
g
H
⁢
θ
1
}
+
ξ
2
⁢

θ
1
-
φ
+
μ
/
ξ

2
s
.
t
.
C
1
:
{
θ
1
H
⁢
q
t
}
≥
ε
3
,
∀
t
C
2
:
❘
"\[LeftBracketingBar]"
θ
1
,
n
❘
"\[RightBracketingBar]"
≤
1
,
∀
n
C
4
:
❘
"\[LeftBracketingBar]"
φ
n
❘
"\[RightBracketingBar]"
=
1
,
∀
n
,
wherein
μ∈N is the dual variable, ξ>0 is a pre-set penalty parameter, and this multi-variable problem is solved by alternatingly updating each variable given the other variables:
updating θ1: given φ as well as μ, an optimization problem regarding θ1 is a convex problem, which is solved by various existing efficient algorithms;
updating φ: given θ1 and μ, φopt is obtained through phase alignment as
φopt=ej∠(ξθ1+μ);
updating μ: given θ1 and φ, the dual variable μ is updated as
μ
:=
μ
+
ξ
⁡
(
θ
1
-
φ
)
;
under given conditions ρ, α, u, w and θ1, an optimization problem regarding a reflection coefficient θ2 is similar to θ1 optimization.

### Claim 9 (depends on 3)

The method for the joint active and passive beamforming and the received signal optimization in the ISAC system assisted by the dual IRSs according to claim 3, wherein the SINR in the S21 is defined as:
a scalar nkCN (0,σk2) is an additive white gaussian noise (AWGN) at the kth user, therefore, the SINR of the kth user is calculated as:
γ
k
=
❘
"\[LeftBracketingBar]"
h
k
H
⁢
w
k
❘
"\[RightBracketingBar]"
2
∑
j
≠
k
K
+
M
❘
"\[LeftBracketingBar]"
h
k
H
⁢
w
j
❘
"\[RightBracketingBar]"
2
+
σ
k
2
,
wherein an effective channel hk∈M is defined as
h
k
H
=
H
k
H
+
F
1
,
k
H
⁢
Θ
1
H
⁢
G
1
+
F
2
,
k
H
⁢
Θ
2
H
⁢
G
2
,
wherein wj denotes a jth column of a beamforming matrix W, i.e., W[w1, . . . , wK+M];
considering a link between the multi-antenna BS and a tth target is blocked, an echo signal received through a path assisted by the IRSs is represented as:
y
r
,
t
=
α
t
(
G
1
H
⁢
Θ
1
⁢
z
1
,
t
+
G
2
H
⁢
Θ
2
⁢
z
2
,
t
)
⁢
(
z
1
,
t
H
,
Θ
1
H
⁢
G
1
+
z
2
H
,
Θ
2
H
⁢
G
2
)
⁢
Ws
+
n
r
;
further considering target detection is performed under far-field conditions, wherein αt represents radar cross section (RCS) of the tth target, i.e., αt=4πP1/S, in which P1 denotes a radiated power density of a target's scattered wave, and S denotes a power density of an incident wave; calculating a sensing target SNR in terms of expectation; for simplifying a formula expression, utilizing σt2 to represent an expected value αt, i.e.,
E
⁢
{
❘
"\[LeftBracketingBar]"
α
t
❘
"\[RightBracketingBar]"
2
}
=
σ
t
2
,
wherein z1,t∈N and z2,t∈N denote baseband channels between IRS1, IRS2 and the tth target, respectively, and a vector
n
r
⁢
▯
⁢
CN
⁢
(
0
,
σ
r
2
⁢
I
M
)
represents the AWGN; assuming the path between the IRSs and a target is LoS, and required angles of arrival/departure (AoA/AoD) are known, an equivalent channel matrix of the echo signal H, is redefined as:
H
t
⁢
▯
⁢
(
G
1
H
⁢
Θ
1
⁢
z
1
,
t
+
G
2
H
⁢
Θ
2
⁢
z
2
,
t
)
⁢
(
z
1
,
t
H
⁢
Θ
1
H
⁢
G
1
+
z
2
,
t
H
⁢
Θ
2
H
⁢
G
2
)
;
defining wvec{W} as vectorizing a matrix W, and ⊗ as a Kronecker product, then yr,t is re-expressed as:
y
r
,
t
=
α
t
(
SS
H
⊗
H
t
)
⁢
w
+
n
r
.

### Claim 10 (depends on 3)

The method for the joint active and passive beamforming and the received signal optimization in the ISAC system assisted by the dual IRSs according to claim 3, wherein the SNR lower bound in the S21 is defined as:
configuring the multi-antenna BS to process a received sensing signal yr,t as:
u
H
⁢
y
r
,
t
=
α
t
⁢
u
H
(
SS
H
⊗
H
t
)
⁢
w
+
u
H
⁢
n
r
,
therefore, an SNR of a tth sensing target is calculated as:
γ
r
,
t
=
σ
t
2
⁢
E
⁢
{
❘
"\[LeftBracketingBar]"
u
H
(
SS
H
⊗
H
t
)
⁢
w
❘
"\[RightBracketingBar]"
2
}
σ
r
2
⁢
u
H
⁢
u
;
for simplifying a formula expression, redefining E{SSH}IK+M; utilizing a Jensen's inequality, i.e., E{ƒ(x)}≥ƒ(E{x}), a resulting SNR lower bound is obtained as:
γ
r
,
t
≥
σ
t
2
⁢
❘
"\[LeftBracketingBar]"
u
H
(
I
K
+
M
⊗
H
t
)
⁢
w
❘
"\[RightBracketingBar]"
2
σ
r
2
⁢
u
H
⁢
u
.
