SM-Bello/ZeroTwin-UAV-Synthetic_Physics-Informed-Multi-UAV-Fault-Telemetry-Benchmark

🛸 ZeroTwin-UAV-Synthetic

Multi-Agent Physics-Informed Degradation Benchmark for Autonomous UAV Swarms

═══════════════════════════════════════════════════════════════════════════════════════ P H I L A B • P E N E L O P E I N C . R E S E A R C H D I V I S I O N ═══════════════════════════════════════════════════════════════════════════════════════

---

🏛️ Provenance & Institutional Trademarks

This open-source benchmark is architected, authenticated, and maintained by PHI Lab, the advanced autonomy and digital twin research unit of Penelope Inc.:

• Research Laboratory: PHI Lab (Physics-Informed Aerospace Systems & Cyber-Physical Intelligence)
• Enterprise Host: Penelope Inc.
• Source Repository: Physics-Hybrid-Integrity-ZeroTwin-Digital-Twin
• Platform Ecosystem: Penelope Digital Twin Core (phi-twin / phi-chain)
• Primary Author: Sonny Bello (Mohammed Bello Sani)
• Distribution Authorization: Penelope Inc. Open Engineering Directive

---

📊 Swarm Telemetry Architecture

---

📐 Differential Physics Engines

The dataset synthesizes continuous multi-channel operational telemetry via four deterministic degradation models:

1. Harmonic Centrifugal Rotor Imbalance

$$F_{\text{imbalance}}(t)=m\cdot r\cdot {\omega }^2\cdot (1+\alpha t)[\sin (\omega t)+0.4\cos (\omega t)]F\_\{\backslash text\{imbalance\}\}(t)\; =\; m\; \backslash cdot\; r\; \backslash cdot\; \backslash omega^2\; \backslash cdot\; (1\; +\; \backslash alpha\; t)\; \backslash left[\backslash sin(\backslash omega\; t)\; +\; 0.4\backslash cos(\backslash omega\; t)\backslash right]$$

• Physical Phenomenon: Dynamic mass displacement yielding structural resonance at fundamental rotational frequency ($50\text{ Hz}$).

2. Lumped-Parameter RC Thermal Dissipation (ESC MOSFETs)

$$T(t)=T_{\text{ambient}}+(T_{\text{steady\_state}}-T_{\text{ambient}})(1-e^{-\frac{t}{{\tau }_{\text{thermal}}}})+I^2{R}_{\text{ds(on)}}T(t)\; =\; T\_\{\backslash text\{ambient\}\}\; +\; \backslash left(T\_\{\backslash text\{steady\backslash \_state\}\}\; -\; T\_\{\backslash text\{ambient\}\}\backslash right)\; \backslash left(1\; -\; e^\{-\backslash frac\{t\}\{\backslash tau\_\{\backslash text\{thermal\}\}\}\}\backslash right)\; +\; I^2\; R\_\{\backslash text\{ds(on)\}\}$$

• Physical Phenomenon: Joulean heating and progressive gate-oxide thermal saturation under heavy aerodynamic thrust loads.

3. Ball Pass Frequency Outer-Race (BPFO) Bearing Wear

$$f_{\text{BPFO}}=\frac{N_{\text{balls}}}{2}\cdot f_{\text{shaft}}\cdot (1-\frac{d}{D}\cos \theta )f\_\{\backslash text\{BPFO\}\}\; =\; \backslash frac\{N\_\{\backslash text\{balls\}\}\}\{2\}\; \backslash cdot\; f\_\{\backslash text\{shaft\}\}\; \backslash cdot\; \backslash left(1\; -\; \backslash frac\{d\}\{D\}\; \backslash cos\; \backslash theta\backslash right)$$

• Physical Phenomenon: Localized spalling defects generating high-frequency acoustic emission ($\text{dB}$) bursts and structural chatter ($g$).

4. Electrochemical Internal Impedance & OCV Decay

$$V_{\text{terminal}}(t)=V_{\text{OCV}}\cdot e^{-\frac{t}{{\tau }_{\text{discharge}}}}-I_{\text{load}}\cdot (R_{\text{int\_0}}+\beta t)V\_\{\backslash text\{terminal\}\}(t)\; =\; V\_\{\backslash text\{OCV\}\}\; \backslash cdot\; e^\{-\backslash frac\{t\}\{\backslash tau\_\{\backslash text\{discharge\}\}\}\}\; -\; I\_\{\backslash text\{load\}\}\; \backslash cdot\; \backslash left(R\_\{\backslash text\{int\backslash \_0\}\}\; +\; \backslash beta\; t\backslash right)$$

• Physical Phenomenon: Anode polarization, cell aging, and ohmic voltage drop under high-current swarm maneuvers.

---

📁 Repository Layout

ZeroTwin-UAV-Synthetic/
├── datasheet.json            # Machine-readable schema & physics specs
├── master.csv                # Complete multi-agent corpus (25,000 samples)
├── uav_node_1.csv            # Edge Client 1 [Lagos Node — Nominal Cruise Bias]
├── uav_node_2.csv            # Edge Client 2 [Beijing Node — Rotor Imbalance Bias]
├── uav_node_3.csv            # Edge Client 3 [Shanghai Node — ESC Overheat Bias]
├── uav_node_4.csv            # Edge Client 4 [Tokyo Node — Bearing Fault Bias]
└── uav_node_5.csv            # Edge Client 5 [Seoul Node — Voltage Sag Bias]

Sensor Schema Dictionary

Attribute	Format	Engineering Unit	Analytical Scope
timestamp	float64	Seconds ($s$)	Flight timeline at uniform $10\text{ Hz}$ resolution
uav_id	int64	Integer ($1 \dots 5$)	Unique swarm agent physical hardware node index
vibration_g	float64	Gravitational Acc ($g$)	Accelerometer stream tracking harmonic imbalance and spalling
temperature_c	float64	Celsius ($^\circ\text{C}$)	ESC MOSFET junction temperature under current loading
voltage_v	float64	Volts ($V$)	Real-time 4S LiPo bus output tracking discharge sag
acoustic_db	float64	Decibels ($\text{dB}$)	Acoustic emission spectrum tracking bearing micro-impacts
fault_label	int64	Index ($0 \dots 4$)	Ground truth integer diagnostic category
fault_name	string	Categorical String	Canonical aerospace fault taxonomy identifier

---

⚡ Quickstart Integration

Pandas Stream Loader

import pandas as pd

# Load verified corpus from Penelope Inc. / PHI Lab repository
REMOTE_PATH = "https://huggingface.co/datasets/<YOUR_HF_USERNAME>/ZeroTwin-UAV-Synthetic/raw/main/master.csv"
corpus = pd.read_csv(REMOTE_PATH)

# Verify per-node Non-IID skew
print(pd.crosstab(corpus['uav_id'], corpus['fault_name']))

PyTorch Dataset Wrapper (Federated-Ready)

import torch
from torch.utils.data import Dataset, DataLoader
import pandas as pd

class PHISwarmDataset(Dataset):
    def __init__(self, filepath: str, seq_len: int = 16):
        raw = pd.read_csv(filepath)
        self.x = torch.tensor(
            raw[['vibration_g', 'temperature_c', 'voltage_v', 'acoustic_db']].values,
            dtype=torch.float32
        )
        self.y = torch.tensor(raw['fault_label'].values, dtype=torch.long)
        self.seq_len = seq_len

    def __len__(self):
        return len(self.x) - self.seq_len

    def __getitem__(self, i):
        return self.x[i : i + self.seq_len], self.y[i + self.seq_len]

# Stream Node 2 for decentralized local training
node2_loader = DataLoader(PHISwarmDataset("uav_node_2.csv"), batch_size=32, shuffle=True)

Hugging Face datasets Loader

from datasets import load_dataset

ds = load_dataset("<YOUR_HF_USERNAME>/ZeroTwin-UAV-Synthetic")
print(ds)

---

🔗 Related Work

• Source Code / Digital Twin Architecture: Physics-Hybrid-Integrity-ZeroTwin-Digital-Twin (GitHub)
• Enterprise Portfolio: Penelope Inc.

---

📜 Intellectual Property, Licensing & Citation

Distributed globally under the Creative Commons Attribution 4.0 International (CC-BY-4.0) license. Maintained by PHI Lab at Penelope Inc.

@dataset{bello2026zerotwin_dataset,
  author       = {Bello, Mohammed Bello Sani and {PHI Lab Research Division}},
  title        = {ZeroTwin-UAV-Synthetic: Physics-Informed Multi-UAV Fleet Degradation Benchmark},
  institution  = {Penelope Inc.},
  year         = {2026},
  publisher    = {Hugging Face},
  url          = {https://huggingface.co/datasets/<YOUR_HF_USERNAME>/ZeroTwin-UAV-Synthetic},
  note         = {Source: https://github.com/Sm-bello/Physics-Hybrid-Integrity-ZeroTwin-Digital-Twin}
}

═══════════════════════════════════════════════════════════════════════════════════════ © 2026–2027 PENELOPE INC. • PHI LAB • ALL RESEARCH TRADEMARKS REGISTERED ═══════════════════════════════════════════════════════════════════════════════════════