A publication-oriented discrete-time simulation framework for evaluating quantum-assisted / quantum-inspired digital twin control architectures in Integrated Sensing and Communication (ISAC) systems for 6G Open RAN networks.
_{Designed for comparative algorithmic evaluation, reproducible experiments, and research prototyping. Not intended to represent a full standards-compliant physical-layer or deployment-grade Open RAN stack.}

---

🏛️ Architecture

Digital Twin Module

Quantum-Assisted Engine

Security Module

Closed-Loop Control Flow

ISAC Signal Model

O-RAN Architecture Alignment

Mapping of framework components to O-RAN functional entities: the digital twin orchestrator sits in the non-RT RIC as an rApp, while the ISAC resource allocation and security monitoring operate as xApps in the near-RT RIC, communicating via standardised A1 (policy) and E2 (control) interfaces.

---

✨ Key Features

Module	Description
6G Open RAN	4 BS × 64 antennas, 40 UE, 10 sensing targets, 1 km² area
Channel	3GPP path loss, log-normal shadowing, Rician fading, AR(1) temporal correlation
Communication	OFDMA downlink, per-RB SINR, Shannon throughput, outage detection
Sensing	Mono-static OFDM radar, Swerling-I P_d, CRLB range estimation, clutter
Digital Twin	Imperfect: sync delay, measurement noise, SINR estimation error, state staleness
Security	EWMA anomaly detection (dual-channel), Bayesian trust dynamics, jamming + spoofing
Quantum Assist	Quantum-inspired candidate screening for uncertainty-aware action ranking
Controller	Closed-loop adaptive: per-BS power control, sensing/comm split, trust gating
Formal Analysis	Trust-gated resource bounds, utility degradation under twin delay, convergence
O-RAN Mapping	Architectural alignment with near-RT RIC / non-RT RIC, A1/E2 interfaces
Evaluation	7 baselines, ablation study, 50 MC runs, 3000 slots, 12+ sweeps, 95% CI

---

⚠️ Current Limitations

This framework is intentionally abstraction-driven and should be interpreted as a research simulation platform rather than a full deployment stack.

• The quantum module is quantum-inspired / quantum-assisted at the algorithmic level, not a hardware-backed quantum optimisation engine
• The sensing model is simplified and intended for comparative ISAC evaluation
• The interference and scheduling models are abstracted for tractability
• Security metrics are scenario-driven and should be interpreted as resilience indicators rather than full intrusion-detection benchmarks

---

📊 Results

Multi-Objective Performance

The Full Proposed method achieves the best overall trade-off across communication, sensing, security, and energy efficiency.

Example Results (illustrative configuration: 3 MC × 300 slots)

Method	Sum Rate (Mbps)	P_d	Trust	Energy	Utility
Static ISAC	1022 ± 110	0.600	1.000	0.996	0.265
Adaptive ISAC	881 ± 95	0.559	1.000	0.997	0.245
DT (no QA)	968 ± 72	0.546	0.839	0.997	0.344
DT+QA (no Sec)	1070 ± 120	0.524	1.000	0.998	0.270
Full Proposed	1082 ± 90	0.743	0.837	0.965	0.366
Uncertainty-Aware	962 ± 98	0.580	1.000	0.997	0.264
UCB Learning	635 ± 85	0.550	1.000	0.998	0.230

Time Evolution — Full Proposed Method

Cumulative Distribution Functions

---

🔬 Sensitivity Analysis

Attack Intensity Sweep

The Full Proposed method maintains superior utility under increasing anomaly probability, demonstrating the value of security-aware closed-loop control.

Digital Twin Delay Sweep

Performance degrades gracefully as twin synchronisation delay increases, validating the framework's robustness to imperfect state information.

---

📐 Formal Analysis

Two analytical results provide guarantees on closed-loop behaviour:

Proposition 1 (Trust-Gated Resource Bound): Under anomaly rate $p_a$ and decay $\beta$, the steady-state trust satisfies $\tau_\infty \geq \alpha / (\alpha + \beta \cdot p_a \cdot \mathbb{E}[\text{excess}])$, ensuring bounded resource gating.

Proposition 2 (Utility Degradation under Twin Delay): For twin sync delay $\delta$ and state decay $\gamma$, the utility loss is bounded by $\Delta J \leq (w_c + w_s)(1 - \gamma^\delta)$, growing monotonically with $\delta$.

Run python main.py --analysis to compute bounds for any configuration.

---

🧪 Ablation Study

Run python main.py --ablation to systematically disable components of the Full Proposed method and measure their individual contribution to overall utility.

Configuration	What is disabled
Full Proposed	Nothing (reference)
No QA	Quantum-assisted candidate search
No Security	Anomaly detection and trust gating
No Twin Adaptation	Twin staleness and sync delay
No Power Adaptation	Online weight tuning
No Sensing Adaptation	Sensing power fraction control

---

🏗️ Baselines

ID	Method	Digital Twin	Quantum Assist	Security	Description
0	Static ISAC	✗	✗	✗	Fixed equal allocation, no adaptation
1	Adaptive ISAC	✗	✗	✗	Measurement-based rebalancing
2	DT-guided	✓	✗	✓	Twin-predicted PF allocation
3	DT+QA (attack-unaware)	✓	✓	✗	Quantum search, blind to attacks
4	Full Proposed	✓	✓	✓	Complete closed-loop system
5	Uncertainty-Aware	✗	✗	✗	SINR-variance-driven robust allocation
6	UCB Learning	✗	✗	✗	Online bandit-based RB optimisation

---

🚀 Quick Start

# Clone
git clone https://github.com/YassirALKarawi/qdt-isac-6g.git
cd qdt-isac-6g

# Install
pip install -r requirements.txt

# Quick test (3 MC, 300 slots)
python main.py --quick

# Full run (50 MC, 3000 slots)
python main.py --mc 50 --slots 3000

# Single baseline
python main.py --baseline 4 --mc 10 --slots 1000

# Parameter sweep
python main.py --sweep anomaly_prob
python main.py --sweep twin_delay
python main.py --sweep clutter

# Ablation study (disables components of Full Proposed one at a time)
python main.py --ablation --mc 10 --slots 1000

# Formal analysis (trust bounds, utility degradation bounds)
python main.py --analysis

---

🔄 Reproducibility

• Python version: 3.9+
• Seed-controlled execution via SimConfig(seed=...)
• Monte Carlo averaging supported via --mc
• Steady-state aggregation computed over the latter 50% of the simulated time horizon
• Slot-level and run-level results exported to CSV under results/
• Figures saved under figures/
• Tests: python -m pytest tests/ -v

---

📐 Composite Utility

$$J = w_c \cdot R_{\text{norm}} + w_s \cdot S + w_{\text{sec}} \cdot T - w_e \cdot E_{\text{norm}}$$

Symbol	Component	Source
$R_{\text{norm}}$	Normalised sum-rate	communication.py
$S$	Sensing utility (P_d + tracking)	sensing.py
$T$	Trust score (Bayesian adaptive)	security.py
$E_{\text{norm}}$	Energy consumption (PA + DSP + circuit)	network.py

Weights adapt online via the closed-loop controller based on observed outage, detection, and trust levels.

---

📁 Project Structure

qdt-isac-6g/
├── config.py            # Simulation parameters + 12 sweep configs
├── channel.py           # Path loss, shadowing, Rician fading, AR(1)
├── network.py           # BS, UE, targets + mobility + energy model
├── communication.py     # SINR, throughput, outage (OFDMA)
├── sensing.py           # Radar SNR, Swerling-I P_d, CRLB
├── digital_twin.py      # Imperfect DT: delay, noise, staleness
├── security.py          # EWMA detection + Bayesian trust
├── quantum_assist.py    # Quantum-inspired candidate search
├── controller.py        # 7 baseline controllers (incl. UCB, uncertainty-aware)
├── simulator.py         # Discrete-time simulation engine
├── analysis.py          # Formal bounds: trust gating, utility degradation, convergence
├── oran_mapping.py      # O-RAN architecture alignment (RIC, A1/E2, xApps)
├── metrics.py           # CSV/JSON metrics export + CI95
├── plotting.py          # Publication-quality figures
├── main.py              # CLI entry point (baselines, sweeps, ablation, analysis)
├── tests/               # 33 verification tests (pytest)
├── figures/             # Generated plots
├── results/             # Output CSV/JSON
├── paper/               # LaTeX manuscript + compiled PDF
├── requirements.txt
├── requirements-dev.txt
└── LICENSE

---

🔧 Available Sweeps (12 experiments)

Sweep	Parameter	Values
user_density	n_users	10, 20, 40, 60, 80
anomaly_prob	anomaly_prob	0.0 – 0.20
twin_delay	twin_sync_delay_slots	0 – 50
clutter	clutter_to_noise_ratio_db	0 – 15 dB
target_speed	target_speed_range	1 – 100 m/s
sensing_power	sensing_power_fraction	0.05 – 0.40
target_density	n_targets	2 – 40
scalability	n_bs	2 – 16
quantum_onoff	qa_enabled	True / False
twin_fidelity	twin_sinr_noise_std	0.5 – 10 dB
mobility	user_speed_range	pedestrian – vehicular
weight_sweep	weight_comm	0.1 – 0.7

---

📖 Citation

@article{alkarawi2026qdt,
  title   = {Quantum-Assisted Digital Twin for Closed-Loop Secure and 
             Adaptive ISAC in 6G Open RAN},
  author  = {Al-Karawi, Yassir Ameen Ahmed},
  journal = {Submitted to IEEE Journal on Selected Areas in Communications},
  year    = {2026},
  note    = {Under review}
}

---

📜 License

The simulation code in this repository is licensed under the MIT License — see LICENSE for details.

The manuscript in paper/ is © 2026 the author. If accepted, copyright will transfer to IEEE per their publication agreement. The manuscript is included for reference only and may not be redistributed without permission.