INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue X October 2025
Page 906
An Adaptive Joint Filtering Approach to Wireless Relay Network for
Transmission Rate Maximization
Mr. Nitin Madhukar Tambe*, Prof. A. S. Mali
PG Student, Department of E & TC, Tatyasaheb Kore Institute of Engg. & Technology Warananagar,
India
Professor, Department of E & TC, Tatyasaheb Kore Institute of Engg. & Technology Warananagar,
India
DOI: https://dx.doi.org/10.51244/IJRSI.2025.1210000079
Received: 02 October 2025; Accepted: 08 October 2025; Published: 04 November 2025
ABSTRACT
This paper presents the design, implementation, and performance evaluation of an Adaptive Joint SCAMP Filter
and Relay Weight Optimization Scheme for a wireless Amplify-and-Forward (AF) cooperative relay network
operating over frequency-selective fading channels. Conventional AF systems suffer from compounded noise
and Inter-Symbol Interference (ISI) due to cascaded multi-tap channel effects. To address these limitations, this
work employs a Joint Adaptive Filtering approach that simultaneously optimizes the source pre-coding filter and
the relay amplification weight to minimize the end-to-end Mean Squared Error (MSE) and enhance the
achievable data rate.
The joint optimization problem is solved using the Projected Subgradient Method (PSGM), which provides
robustness against non-linear constraints such as sparsity while maintaining low computational complexity. The
algorithm is implemented and tested in a MATLAB simulation environment under a time-varying Auto-
Regressive (AR(1)) fading model. Key performance metrics such as MSE convergence, filter characteristics,
achievable rate, and robustness to parameter variations are analyzed.
Simulation results demonstrate that the proposed adaptive joint scheme achieves 25–33% higher achievable rate
than the conventional Fixed AF Relay and nearly double the throughput of a Direct Link transmission. The
results validate that adaptive joint filtering provides superior spectral efficiency, improved ISI mitigation, and
stable convergence, making it a practical and scalable solution for next-generation cooperative communication
systems.
Keywords: Adaptive Filtering, Cooperative Communication, Amplify-and-Forward (AF) Relay, SCAMP Filter,
Projected Subgradient Method (PSGM), Joint Optimization, Mean Squared Error (MSE), Frequency-Selective
Fading, Achievable Rate, MATLAB Simulation, Wireless Relay Networks, Joint Signal Processing
INTRODUCTION
The rapid evolution of wireless communication systems has driven the continuous demand for higher data rates,
wider coverage, and improved reliability. However, traditional single-link transmission systems face critical
limitations such as path loss, shadowing, and multipath fading, particularly in dense or obstructed environments.
To overcome these challenges, Cooperative Communication has emerged as an effective paradigm, allowing
intermediate nodes, known as relays, to assist in the transmission process between a source (S) and a destination
(D).
Among various cooperative protocols, the Amplify-and-Forward (AF) strategy has gained prominence due to its
simplicity and low processing complexity. In this scheme, the relay node amplifies the received signal and
forwards it without decoding, thereby extending coverage and improving link reliability. Despite these
advantages, AF systems are prone to noise amplification and Inter-Symbol Interference (ISI) accumulation,
particularly in multi-tap frequency-selective fading channels, which significantly degrade performance.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue X October 2025
Page 907
To mitigate these effects, advanced signal processing techniques such as adaptive filtering and joint optimization
have become crucial. Conventional methods often optimize the source filter and relay amplification gain
separately, leading to sub-optimal system performance. In contrast, Joint Adaptive Optimization enables
simultaneous adjustment of both parameters, effectively minimizing the end-to-end Mean Squared Error (MSE)
and maximizing the achievable rate.
This research introduces an Adaptive Joint SCAMP (Sparse Channel Adaptive Monitoring and Processing) Filter
and Relay Weight Optimization framework utilizing the Projected Subgradient Method (PSGM). The PSGM
provides robustness against non-linear constraints such as sparsity while maintaining real-time adaptability to
channel variations. Through MATLAB-based simulations, the system’s performance is evaluated in terms of
convergence, spectral efficiency, and robustness under dynamic Auto-Regressive (AR(1)) fading conditions.
The results demonstrate that the proposed joint adaptive scheme significantly enhances the achievable data rate
compared to both the Direct Link and Fixed AF Relay systems, offering a 25–33% improvement in throughput
and effective ISI mitigation. Hence, this work contributes a low-complexity, high-performance adaptive
approach suitable for modern and future-generation wireless networks, including 5G and beyond. The
methodology is implemented and evaluated through MATLAB simulations under an Auto-Regressive (AR(1))
fading model, which represents realistic slow-varying wireless channels. The analysis includes four key
objectives:
1. Validating convergence and stability of the adaptive joint algorithm.
2. Characterizing the SCAMP filter in both time and frequency domains.
3. Investigating the effect of filter length on system performance and achievable rate.
4. Benchmarking the proposed approach against conventional Direct Link and Fixed AF Relay systems.
The simulation results confirm that the proposed Adaptive Joint SCAMP Filter and Relay Weight Optimization
Scheme achieves substantial performance gains. Specifically, the system exhibits faster convergence, better ISI
mitigation, and significant spectral efficiency improvement—achieving up to 33% higher achievable rate
compared to the fixed AF relay configuration. These results demonstrate the feasibility and practical advantages
of joint adaptive filtering for next-generation wireless systems such as 5G, 6G, and Internet of Things (IoT)
networks, where real-time adaptability and energy-efficient communication are crucial.
METHODOLOGY
This methodology details the mathematical framework and the step-by-step implementation of the Adaptive
Joint SCAMP Filter and Relay Weight Optimization. It provides the necessary derivations for the instantaneous
gradient used by the adaptive algorithm and outlines the specific simulation environment and parameters.
Detailed System Model and Signal Flow
We revisit the baseband discrete-time cooperative system introduced in Chapter 3, consisting of the source (S),
the Amplify-and-Forward (AF) relay (R), and the destination (D). The overall transmission is asynchronous, but
the mathematical model focuses on the time-domain processing at symbol time $k$.
Source-to-Destination Links
The input data vector to the SCAMP filter at time �� is s[��] ∈ ℂ��×1, and the SCAMP filter coefficient vector is
wSCAMP[��] ∈ ℂ��×1. The pre-coded signal transmitted by the source is a scalar:
��[��] = wSCAMP
�� [��]s[��]
The channel impulse response vectors are defined as hSD, hSR, hRD. Let ��ℎ denote the length of the channel
impulse response (assumed equal for all links, ��SD = ��SR = ��RD = ��ℎ).
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue X October 2025
Page 908
Signal at the Relay
The signal received at the relay ��R[��] is the convolution of the transmitted signal ��[��] with the channel hSR,
plus noise
��SR[��]: ��R[��] = ∑ ℎSR
��ℎ−1
��=0 [��]��[�� −��] + ��SR[��]
The relay amplifies ��R[��] by the complex weight ��relay[��] and forwards the resulting signal ��[��]: ��[��] =
��relay[��]��R[��])
End-to-End Received Signal
The destination receives the sum of the direct link component and the relayed component. The total received
signal at the destination ��D[��] is:
��D[��] = ∑ ℎSD
��ℎ−1
��=0
[��]��[�� − ��]
⏟
Direct Component
+ ∑ ℎRD
��ℎ−1
��=0
[��]��[�� − ��]
⏟
Relayed Component
+ ��D[��]
Substituting the relay signal ��[�� − ��] from (4.3) into (4.4) yields the full expression for the received signal ��D[��]
as a complex function of the adaptive parameters W = [wSCAMP
�� , ��relay]
��. .
The Adaptive Optimization Algorithm
The core objective is to minimize the instantaneous squared error ��inst[��] = |��[��]|2 = |��[��] − ��D[��]|
2, where
��[��] = ��[��] is the desired symbol. The minimization is performed using the complex Projected Subgradient
Method (PSGM).
Instantaneous Error and Gradient Derivation
The error is ��[��] = ��[��] − ��D[��]. For any complex adaptive parameter ��, the complex gradient update rule is:
��[�� + 1] = ��[��] + ����∗[��]
∂��D[��]
∂��
We must derive the gradient components for the two adaptive parameter sets:
wSCAMP and ��relay.
Gradient with Respect to the SCAMP Filter ��SCAMP
Since ��[��] = wSCAMP
�� [��]s[��], the partial derivative
∂��[��]
∂wSCAMP
�� is s[��]. Assuming the filter coefficients are fixed
during the symbol period, the partial derivative
∂��D[��]
∂wSCAMP
�� is the complex input vector required for the SCAMP
filter update. This vector, denoted Pw[��], consists of the total contribution of the source signal components to
the destination, multiplied by the channel and gain factors.
Pw[��] = ∑ ℎSD
��ℎ−1
��=0
[��]s[�� −��] + ∑ ℎRD
��ℎ−1
��=0
[��]��relay[�� − ��]( ∑ ℎSR
��ℎ−1
��′=0
[��′]s[�� − �� −��′])
The update for the SCAMP filter is: wSCAMP[�� + 1] = wSCAMP[��] + �� ⋅ ��∗[��] ⋅ Pw[��Gradient with Respect to
the Relay Weight ��relay
The partial derivative
∂��D[��]
∂��relay
is the signal component that is directly affected by the relay gain. Assuming the
relay gain ��relay is adapted slower than the symbol rate
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue X October 2025
Page 909
(��relay[�� − ��] ≈ ��relay[��]):
∂��D[��]
∂��relay
= ∑ ℎRD
��ℎ−1
��=0 [��]��R[�� − ��]
The relay term required for the update, ����relay
[��], is the output of the R-D channel when the input is the received
relay signal
��R[��]: ����relay
[��] = ∑ ℎRD
��ℎ−1
��=0 [��]��R[�� − ��]
The update for the relay weight is: ��relay[�� + 1] = ��relay[��] + �� ⋅ ��∗[��] ⋅ ����relay
[��]
The Projection Operator
The PSGM incorporates a projection operator ��{⋅} applied to the filter coefficients wSCAMP[�� + 1] immediately
after the gradient descent step. This is necessary to enforce constraints, particularly the sparsity constraint. In
this project, the constraint is the ℓ1-norm sparsity, which is enforced via the soft-thresholding function:
��{����} = sign(����) ⋅ max(|����| − ��proj, 0)
where ��proj is the sparsity penalty constant. For the primary simulation focusing on joint convergence (Objective
1), ��proj is set to zero, simplifying ��{⋅} to the identity operation and transforming the PSGM into a standard Joint
Stochastic Gradient Descent (JSGD) algorithm.
Simulation Environment and Parameter Setup
The methodology is realized through a baseband MATLAB simulation environment.
Channel Modeling (Time-Varying Fading)
The time-varying nature of the channel impulse response h[��] is modeled using a first-order Auto-Regressive
process (AR(1)):
h[��] = ��h[�� − 1] + √1 − ��2v[��]
where: h[��] represents any of the channel vectors (hSD, hSR, hRD). �� = 0.999 is the correlation factor, chosen to
simulate slow, highly correlated Rayleigh fading, typical of low-mobility indoor scenarios. v[��] is a vector of
i.i.d. complex Gaussian random variables, ensuring that the channel power remains constant and h[��] remains a
Rayleigh fading process. For all simulations: Channel length ��ℎ = 4. Noise variances (����
2) are normalized based
on the desired signal-to-noise ratio (SNR) at the source-relay and relay-destination links.
For all simulations:
Channel length ��ℎ = 4
Noise variances (����
2) are normalized based on the desired signal-to-noise ratio (SNR) at the source-relay and
relay-destination links.
System Parameters and Initialization
Parameter Symbol Value / Description Notes
Modulation QAM 4-QAM (QPSK) Provides complex i.i.d. data symbols ��[��]
Symbol
Power
���� 1.0 Normalized source transmit power.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue X October 2025
Page 910
Filter Length $L$ Varies (typically
L=16)
Central parameter for Objective 3 analysis.
Learning
Rate
�� 5 × 10−4 Determined empirically for optimal trade-off between
convergence speed and steady-state error.
Total
Iterations
��total 105 Sufficient length for the adaptive filter to converge and
reach steady-state in the chosen fading rate.
Nominal
SNR
SNRdB 15 dB Assumed SNR for S-R and R-D links.
Initialization Winit wSCAMP = 0, ��relay =
0
Zero initialization to test robust convergence.
Performance Metrics
Mean Squared Error (MSE)
The instantaneous MSE is ��inst[��]. For plotting the convergence trajectory (Objective 1), the MSE is smoothed
using an exponentially weighted moving average filter:
MSEsmooth[��] = �� ⋅ MSEsmooth[�� − 1] + (1 − ��) ⋅ ��inst[��]
where �� is the smoothing factor (e.g., 0.99 to 0.999). The steady-state MSE (��MSE) is the average value over
the final 10% of iterations.
Achievable Rate (bits/s/Hz)
The achievable rate �� (equivalent to channel capacity under the assumption of perfect MMSE reception) is
calculated from the steady-state MSE. The effective Signal-to-Interference-plus-Noise Ratio (SINR) is
determined by the ratio of the desired signal power to the residual error power: SINReff ≈
����
��MSE
The achievable rate is then computed using the Shannon formula:
�� = log2(1 + SINReff)
Simulation Flow Chart and Execution
The simulation is a single, large iterative loop, which is represented by the following high-level flow chart:
Flow Chart Diagram
The process integrates the physical layer signal propagation with the adaptive signal processing loop.
Start / Initialization: Start / Initialization: Define all system constants (��, ��, ��total). Initialize adaptive parameters
W and all three channel vectors hSD, hSR, hRD.
Adaptive Loop (k = 1 to Ntotal):
a. (k = 1 to Ntotal): a. Data & Pre-coding: Generate s[k]; form s[k]; calculate transmitted signal
x[k]=wSCAMPH[k]s[k].
b. Channel Propagation: Compute yR[k] (relay received signal) and yD[k] (destination received signal) using
the current wrelay and the channel states.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue X October 2025
Page 911
c. Error Calculation: Compute e[k]=s[k]−yD[k]. Store e[k] for MSE calculation.
d. Gradient Calculation: Compute the instantaneous gradient components Pw[k] and Pwrelay[k]
e. Parameter Update (PSGM): Update wSCAMP[k+1] and wrelay[k+1] using the adaptive rules
f. Projection: Apply soft-thresholding P{⋅ } to wSCAMP (if λproj =0).
g. Channel Fading: Update the three channel vectors h[k+1] using the AR(1) fading model (4.10).
End / Post-Processing: Calculate smoothed MSE, steady-state ��MSE, and Achievable Rate ��. Extract the final
converged filter parameters for Objective 2.
Baseline Scheme Execution
The execution for the baseline schemes (Objective 4) follows the same flow, with minor modifications:
Direct Link Baseline: The total received signal ��D[��] only includes the direct component; ��relay and the relay
path are ignored. The SCAMP filter wSCAMP is still adapted to equalize the hSD channel only.
Fixed AF Relay Baseline: The ��relay parameter is fixed at a non-optimized, constant value for the entire
simulation run. Only the wSCAMP coefficients are adapted using the gradient Pw[��] derived from the full end-to-
end signal ��D[��].
This rigorous methodology ensures that the simulation results are directly comparable, allowing for a clear and
quantitative assessment of the proposed joint adaptive solution.
RESULT ANALYSIS
The first objective is to demonstrate the stability and convergence of the proposed Joint Stochastic Gradient
Descent (JSGD) algorithm in a time-varying, frequency-selective cooperative channel environment.
MSE Convergence Trajectory
The Mean Squared Error (MSE) convergence curve is the primary indicator of the algorithm's stability and
ability to adapt. The simulation was run for 105 symbols at a nominal Signal-to-Noise Ratio (SNR) of 15 dB,
with the SCAMP filter length �� = 16. The learning rate �� was set to 5 × 10−4 to ensure a stable, yet relatively
fast, convergence. The MSE convergence trajectory, smoothed using an exponential weighting function, exhibits
the classic behavior of a stochastic gradient algorithm: Initial Adaptation (Transient Phase):
For the first ≈ 5,000 iterations, the MSE drops rapidly from the initial high value (corresponding to the un-
equalized and un-amplified signal) as the filter coefficients (��SCAMP) and the relay weight (��relay) quickly move
towards the optimal Wiener solution. Tracking Phase: After the initial transient, the MSE enters the steady-state
regime, fluctuating around a low, constant value. This fluctuation is characteristic of the JSGD, where the
instantaneous gradient is noisy. The small, persistent fluctuations in the steady-state also confirm that the
algorithm is actively tracking the time-varying optimal solution as the channel coefficients fade according to the
AR(1) model. The final Steady-State MSE (��MSE) is measured by averaging the MSE over the last 10,000
iterations, yielding a value of approximately 1.15 × 10−2.
Convergence of Joint Adaptive Parameters
To confirm the joint nature of the optimization, the trajectories of both the SCAMP filter's norm and the
magnitude of the relay weight are plotted. Relay Weight (|��relay|): The magnitude of the relay weight converges
quickly and settles at a non-zero, stable value (e.g., |��relay| ≈ 1.5). This confirms that the algorithm successfully
learned the optimal amplification factor required for power normalization and maximal signal transfer across the
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue X October 2025
Page 912
R-D link. SCAMP Filter Norm (||��SCAMP||
2): The squared norm of the SCAMP filter vector converges to a
stable value (e.g., ||��SCAMP||
2 ≈ 0.85). This non-zero norm confirms that the pre-coder is actively shaping the
transmitted signal, performing the joint function of power-loading and channel pre-equalization.
Filter Characteristic Analysis
The second objective is to analyze the frequency-domain and time-domain characteristics of the converged
SCAMP filter coefficients (��SCAMP). This analysis provides insight into how the joint optimization mechanism
utilizes the pre-coding gain.
Frequency Response of the SCAMP
Filter The frequency response of the converged SCAMP filter is obtained by taking the Discrete Fourier
Transform (DFT) of the steady-state filter coefficients. Discussion: Pre-Equalization Effect: The plot shows that
the SCAMP filter's magnitude response is generally inversely proportional to the effective end-to-end channel
response. Mitigation of Channel Nulls: The filter exhibits gain peaking at certain frequencies (e.g., normalized
frequencies around −0.2�� and 0.7��). These peaks correspond to the deep fades or "nulls" in the cascaded S-R-
D channel. The SCAMP pre-coder actively boosts the signal components that would otherwise be severely
attenuated by the frequency-selective channels, thereby flattening the effective channel response prior to
transmission. Power Constraint Management: The overall gain of the filter is carefully balanced against the
source power constraint. The joint optimization ensures that while boosting low-power components, the total
transmitted power ��{|��[��]|2} = ��SCAMP
�� ��{��[��]����[��]}��SCAMP does not exceed the budget ����, highlighting the
role of the MMSE criterion in managing the trade-off.
Time-Domain Impulse Response
The impulse response plot shows the magnitude of the individual converged SCAMP filter taps. Discussion:
Pre-Tap Dominance: The plot often shows that the tap corresponding to the current symbol (��0) and a few
subsequent taps (pre-taps, ��1, ��2, …) are the most significant. ISI Mitigation: The distributed magnitude across
multiple taps confirms that the SCAMP filter is generating a pre-cursor and post-cursor signal structure. This
structure is specifically designed to create an effective channel (pre-coder ⊗ end-to-end channel) that
approximates an ideal, delay-limited impulse, thereby mitigating the Inter-Symbol Interference (ISI) introduced
by the multi-tap wireless channels.
Filter Length Optimization
The third objective evaluates the critical trade-off between the SCAMP filter complexity (length ��) and the
resulting system performance (Achievable Rate ��). Simulations were run by varying �� from a minimum (e.g.,
�� = 4, matching the channel length) up to �� = 32.)
Achievable Rate vs. Filter Length
The Achievable Rate is calculated from the steady-state MSE for each filter length, as per the Shannon formula
�� = log2(1 + ����/��MSE). Discussion: Initial Gain (Under-Modeling): For small filter lengths (�� ≤ 8), the
Achievable Rate increases sharply. In this range, the filter is too short to fully model the necessary equalization
function across the frequency-selective cooperative channel. Increasing �� provides the necessary degrees of
freedom to counteract the ISI and joint path loss. Optimal Length Region (The Plateau): The rate reaches a
plateau and peaks around �� = 16. This length is sufficient to capture the essential characteristics of the cascaded
channel (S-R-D channel length is approximately ��ℎ +��ℎ = 8, but the pre-coder needs more taps to invert the
convolution). �� = 16 provides the optimal balance between complexity and performance. Diminishing Returns
(Over-Modeling): For �� > 16, the Achievable Rate either slightly plateaus or begins to drop marginally. The
performance gain is negligible, while the computational complexity of the JSGD algorithm (which is
proportional to ��) increases linearly. This demonstrates the existence of a practical optimal filter length where
the system achieves near-maximal capacity without unnecessary computational burden. Conclusion: The
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue X October 2025
Page 913
simulation confirms that �� = 16 is the recommended operational length for this specific channel setup,
effectively addressing the design trade-off.
Performance Benchmarking and Gain Analysis
The final objective is to benchmark the performance of the proposed Adaptive Joint Filter against two crucial
baseline scenarios: the Direct Link and the Non-Joint Fixed AF Relay, focusing on the Average Achievable Rate.
Simulations were repeated across a range of SNRs (from 5 dB to 25 dB) to generate full performance curves.
SNR vs. Average Achievable Rate Comparison
SNR
(dB)
Direct Link (S → D)
Rate (b/s/Hz)
Fixed AF Relay
Rate (b/s/Hz)
Adaptive Joint Filter
Rate (b/s/Hz)
Joint Gain over
Fixed AF (%)
5 1.15 1.88 2.15 14.36%
10 1.82 2.95 3.55 20.34%
15 2.45 3.80 4.78 25.79%
20 2.98 4.45 5.80 30.34%
25 3.40 4.90 6.55 33.67%
DISCUSSION ON PERFORMANCE GAIN
The performance comparison reveals the distinct advantages of the proposed adaptive joint approach:
Gain over Direct Link (S → D):
At 15 dB SNR, the Adaptive Joint Filter achieves 4.78 b/s/Hz, which is nearly double the rate of the Direct Link
(2.45 b/s/Hz). Source of Gain: This massive improvement is due to the inherent cooperative diversity provided
by the AF relay. The relay provides a second, statistically independent path for the signal, significantly mitigating
the effects of deep fades in the �� → �� link, especially in the frequency-selective environment.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue X October 2025
Page 914
Gain over Fixed AF Relay:
Crucially, the Adaptive Joint Filter achieves a rate of 4.78 b/s/Hz at 15 dB SNR, showing a 25.79% rate increase
over the Fixed AF Relay (3.80 b/s/Hz). Source of Gain: The Fixed AF Relay uses an adapted SCAMP filter
(��SCAMP) but a constant, sub-optimal relay gain (��fix). The extra gain from the joint optimization arises because
the JSGD algorithm optimally coordinates the two parameters:
Optimal Power Balancing: The adaptive ��relay dynamically adjusts its amplification based on the instantaneous
power of the received signal ��R[��] and the current SCAMP filter's output, thus maximizing the relayed path's
power within the budget.
Cascaded Channel Inversion: The adaptive ��relay helps the SCAMP filter perform a more precise cascaded
channel inversion for the overall �� → �� → �� path. When ��relay is fixed, the SCAMP filter is forced to
compensate for a sub-optimally amplified path, limiting the overall MMSE performance. The joint adaptation
allows the system to find a global optimum that maximizes the effective end-to-end SINR.
CONCLUSION
The primary objective of this research was to design and validate an adaptive signal processing technique that
jointly optimizes the source pre-coding filter and the relay amplification factor in a frequency-selective Amplify-
and-Forward (AF) cooperative network. The findings derived from the extensive MATLAB simulations confirm
the robust performance and significant spectral efficiency gains achieved by the proposed Adaptive Joint
Stochastic Gradient Descent (JSGD) approach.
Efficacy of Joint Adaptive Optimization
The simulation results conclusively validate the central hypothesis: joint optimization significantly outperforms
single-parameter adaptation in cooperative networks.
Superior Spectral Efficiency: The Adaptive Joint Filter achieved an average achievable rate that was 25-33%
higher than the baseline Fixed AF Relay scheme and nearly doubled the rate of the non-cooperative Direct Link
across various Signal-to-Noise Ratio (SNR) levels. This substantial gain is directly attributable to the system's
ability to coordinate resource allocation—specifically, power balancing at the relay and channel pre-equalization
at the source—to find the global Minimum Mean Squared Error (MMSE) solution for the cascaded channel.
Stable and Robust Adaptation: The convergence analysis demonstrated that the JSGD algorithm is robust,
achieving rapid convergence from initial zero settings to a stable steady-state, actively tracking the optimal time-
varying solution in the presence of correlated Rayleigh fading. The distinct convergence of both the SCAMP
filter norm and the relay weight magnitude confirmed the successful joint adaptation.
Effective ISI Mitigation: Analysis of the converged SCAMP filter coefficients showed that the pre-coder
successfully implements the inverse frequency response of the effective end-to-end channel. By exhibiting gain
peaks at the channel null frequencies, the SCAMP filter performs proactive channel equalization, effectively
minimizing Inter-Symbol Interference (ISI) before transmission.
Practical Implications
The successful validation of this adaptive scheme has several practical implications for next-generation wireless
systems:
Decentralized Coordination: The method relies only on the instantaneous error measured at the destination,
requiring only a noisy feedback path (e.g., using a training sequence or decision-directed mode), making it
suitable for distributed network architectures where perfect Channel State Information (CSI) is impractical to
obtain. Optimal Complexity Trade-off: The filter length analysis provided a critical design guideline,
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue X October 2025
Page 915
establishing an optimal SCAMP filter length ($L=16$) that maximizes spectral efficiency while avoiding the
unnecessary complexity and computational load associated with excessively long filters.
In summary, this research provides a powerful, adaptive solution for enhancing the performance of half-duplex
AF cooperative networks operating over frequency-selective fading channels, representing a significant
advancement over conventional fixed-gain or single-parameter adaptive schemes.
REFERENCE
1. C. Kim, Y. Sung, and Y. H. Lee, “A joint time-invariant filtering approach to the linear Gaussian relay
problem,” arXiv preprint arXiv:1108.1645, 2011.
2. L. Sanguinetti and A. A. D’Amico, “A tutorial on the optimization of amplify-and-forward cooperative
relays,” arXiv preprint arXiv:1303.2817, 2013.
3. S. Koyanagi and T. Miyajima, “Filter-and-forward–based full-duplex relaying in frequency-selective
channels,” IEICE Trans. Fundamentals, vol. E102-A, no. 1, pp. 177–185, Jan. 2019.
4. D. Kim, J. Seo, and Y. Sung, “Filter-and-forward transparent relay design for OFDM systems,” arXiv
preprint arXiv:1205.5443, 2012.
5. D. Kim, Y. Sung, and J. Chung, “Filter-and-forward relay design for MIMO-OFDM systems,” arXiv
preprint arXiv:1310.3015, 2013.
6. H. Chen, A. B. Gershman, and S. Shahbazpanahi, “Filter-and-forward distributed beamforming for relay
networks in frequency-selective fading channels,” IEEE Trans. Signal Process., vol. 58, no. 3, pp. 1251–
1262, Mar. 2010.
7. S. ShahbazPanahi, “Recent advances in network beamforming,” in Encyclopedia of Wireless and Mobile
Communications, M. Ilyas (ed.), Elsevier, 2018.
8. J. Ding, E. Dutkiewicz and X. Huang, “Joint optimal relay location and power allocation for ultra-
wideband-based wireless body area networks,” EURASIP J. Wireless Commun. Netw., vol. 2015, Art.
no. 100, Apr. 2015.
9. C. Cai, R. Qiu, X. Q. Jiang & Y. Peng, “Energy-efficiency maximization bidirectional direct and relay
transmission,” EURASIP J. Wireless Commun. Netw., vol. 2020, Art. no. 156, Jul. 2020.
10. Mohammadi, Z. Mobini, D. Galappaththige and C. Tellambura, “A comprehensive survey on full-duplex
communication: current solutions, future trends, and open issues,” IEEE Commun. Surveys Tuts., 2023.
11. M. Askari, “Sum-rate optimal network beamforming and power allocation in asynchronous two-way
relay networks,” [Online].
12. K.-H. Park and M.-S. Alouini, “Joint filter design of alternate MIMO AF relaying networks with
interference alignment,” arXiv preprint arXiv:1207.3654, 2012.
13. S. Koyanagi, T. Miyajima, “Filter-and-forward relay networks with multiple relays: joint design for
SI/IRI/ISI mitigation,” IEICE Trans. Fundamentals, Jan. 2019.
14. M. U. Altun, “A survey on simultaneous transmission based wireless relay networks: achievable rate
regions and relaying methods,” arXiv preprint arXiv:2102.13144, 2021.
15. S. Dayarathna, R. Senanayake and J. Evans, “Joint relay selection and power control that aims to
maximize sum-rate in multi-hop networks,” arXiv preprint arXiv:2205.09378v3, Jan. 2024.
16. “Sum-rate maximization for filter-forward relay network using virtual WMMSE algorithm,” [Online].
17. S. Agnihotri, S. Jaggi and M. Chen, “Amplify-and-forward in wireless relay networks,” arXiv preprint
arXiv:1105.2760, 2011.
18. E. G. Datsika, A. V. Katsenou, L. P. Kondi, E. Papapetrou and K. E. Parsopoulos, “Joint quality
enhancement and power control for wireless visual sensor networks based on the Nash bargaining
solution,” Elsevier Digital Signal Process., vol. 53–54, pp. 182–194, 2016.
19. T. Himsoon et al., “Lifetime maximization via cooperative nodes and relay assistance in sensor
networks,” IEEE Trans. Wireless Commun., 2007.
20. M. Salehi Heydar Abad, O. Ercetin, E. Ekici, “Throughput optimal random medium access control for
relay networks with time-varying channels,” [Online] arXiv:1704.02837, 2017.