AETHERIX

Autonomous Extraterrestrial High-throughput Enhancing Routing and Inter-planetary eXchange

EduQual Level 6 — Topic 59

Building an Interplanetary Communication Network with DTN, Quantum Communication, and Space-Based Infrastructure for Mars Mission Support

Muhammad Abdullah Tariq

September 2026

Live Showcase Source Code

Agenda What We'll Cover

[A2] topology.py · References at slides 25–26 · ~18 minutes

Overview What is AETHERIX & Why It Matters

What is AETHERIX?

• An architecture for delay-tolerant networking (DTN) between Earth and Mars
• Replaces static Contact Graph Routing with reinforcement-learning routing
• Secures links with Quantum Key Distribution (BB84 & E91 + repeaters)
• Hybrid optical/RF links across a 5-tier, 241-node topology
• Working proof-of-concept: 33 Python modules, 643 tests, 15 browser demos

The Problem

• TCP/IP breaks — 3–22 min one-way delay; 6–44 min round trip
• Current Mars links: 0.5–6 Mbps, 60–75% availability [1]
• Solar conjunction: ~2-week blackout every 26 months
• Static routing can't adapt to dynamic contact schedules
• Classical key exchange is infeasible across interplanetary distance

AETHERIX addresses every one: DTN store-and-forward, AI-driven routing, quantum-secured keys, and hybrid optical/RF links with Lagrange-relay conjunction coverage.

[1] NASA MRO · [12] RFC 4838 · [A2] topology.py

The Problem Why Space Breaks the Internet

Earth → Mars

Closest approach	54.6 million km
Farthest distance	401 million km
Distance range	7.3× variation
One-way light time	3 – 22 minutes
Synodic period	779.94 days (26 months)

Why TCP/IP Fails

TCP/IP expects	Deep-space reality
RTT < 200 ms	6 – 44 min RTT
Always connected	Scheduled contacts only
End-to-end path	No simultaneous route
ACKs in seconds	ACKs take minutes–hours

“TCP/IP assumes the network is fast, reliable, and always connected. Deep space is none of those things.”

[3] JPL Horizons (distance/light-time) · [12] RFC 4838 · [1] NASA MRO

Solution Delay-Tolerant Networking — The Postal Service of Space

Bundle

→

Store

→

Wait

→

Forward

→

Deliver

BPv7

Store & forward with custody (RFC 9171)

RL

Multi-agent federated Q-learning

QKD

BB84 & E91 + repeater chains

Each node takes custody of the bundle and forwards it one hop closer when a contact window opens — no continuous end-to-end connection is ever required.

[9] RFC 9171 BPv7 · [10] RFC 5326 LTP · [11] RFC 7242 TCPCL · [12] RFC 4838

Architecture System Architecture

[A1] rl_agent.py · [A2] topology.py · [A4] link_budget.py · [A5] qkd.py · github.com/matx104/AETHERIX

Protocol Bundle Protocol v7 — The Foundation

Protocol Stack

• Application — science data, commands, telemetry
• Bundle Protocol v7 — custody, fragmentation, TTL, priority
• Convergence — LTP (deep space), TCPCL (Earth), UDP-CL (optical ISL)
• Physical — optical 1550 nm / Ka-band RF

Custody transfer = each node takes legal responsibility, then the previous custodian frees its buffer.

Priority Classes (P0–P4)

P0 EMERGENCY	Life-critical alerts, abort
P1 URGENT	Nav corrections, hazards
P2 NORMAL	Telemetry, science, commands
P3 LOW	Software updates, bulk files
P4 BULK	Background sync, archival

Standards: [9] RFC 9171 · [12] RFC 4838 · [10] RFC 5326 LTP · [2] CCSDS 734.2-B-1

Protocol Protocol Stack — BPv7 over Three Convergence Layers

[9] RFC 9171 · [10] RFC 5326 · [11] RFC 7242 · [2] CCSDS 734.2-B-1

Architecture Five-Tier Network — 241 Nodes Across Two Worlds

Tier Breakdown (matches topology.py)

Tier	Nodes	Assets
T1 Earth Ground	5	3 DSN (Goldstone, Madrid, Canberra) + MOC + SOC
T2 Earth Orbital	51	3 GEO + 48 LEO laser mesh
T3 Deep Space	4	ES-L4, ES-L5 + 2 transfer-orbit relays
T4 Mars Orbital	4	2 areostationary + 1 polar + 1 orbiter
T5 Mars Surface	177	2 bases, 5 rovers, 10 drones, 160 sensors
Total	241

Link Characteristics

Link	Rate	Type
Earth ↔ Deep Space	100 Mbps	1550 nm optical
Deep Space ↔ Mars	2–200 Mbps	distance-dependent
Mars Orbit ↔ Surface	2 Mbps	UHF / X-band
LEO inter-satellite	10 Gbps	optical ISL

3 redundant paths · no single point of failure · Lagrange relays for conjunction coverage.

[A2] topology.py (241 nodes, 5 tiers) · [1] NASA DSN · [3] JPL Horizons

Network 5-Tier Network Diagram

[A2] topology.py · [3] JPL Horizons (ES-L4/L5 geometry) · [1] NASA DSN

Link Budget Optical Communications — 1550 nm

100–200

Mbps at closest approach

10–20

Mbps at average distance

2–5

Mbps at farthest point

Rates above are capability targets for a coded DSOC-class link [4], not measured throughput.

Key Equations

• FSPL = 20·log₁₀(4πd/λ)
• Gain = 10·log₁₀(η·(πD/λ)²)
• Pᵣ = Pᵗ + Gᵗ + Gᵣ − FSPL

Rate degrades with distance squared — 10–100× capability over the RF baseline (target).

Computed margins [A4] are a conservative uncoded photon-counting budget: +4.31 / +3.33 / +3.08 dB at 55M / 225M / 401M km. It excludes SCPPM coding gain and high-order PPM photon efficiency (≈13 dB), so it is a floor, not a rate prediction.

Configuration

Wavelength	1550 nm (telecom heritage)
Tx power	5 W
Tx aperture	22 cm
Rx aperture	5.0 m (Hale-class ground)

[5] CCSDS 141.0-B-1 · [A4] link_budget.py (physics-derived) · [1] NASA MRO (6 Mbps RF baseline) · [4] NASA DSOC

Data Flow End-to-End Bundle Journey — Mars to Earth

[A2] topology.py (full data path) · [3] JPL Horizons (light-time) · delivery ratio is a design target

AI Reinforcement-Learning Routing

Contact Graph Routing (baseline)

• Pre-computed contact plans required
• Cannot adapt to unexpected link failures
• Re-plans on a stale schedule
• No learning from historical performance

AETHERIX RL Agent

• State: node, neighbours, link quality, buffer
• Actions: forward | store | drop | split
• Reward: R = α(delivery) − β(delay) − γ(hops) − δ(drops) − ε(energy)
• ε-greedy, decay 0.995; α=1.0, δ=10.0 (drops forbidden)
• Q-table + federated aggregation — every value auditable

713/800

forward decisions [A3]

seconds

recovery vs hours

Q-table

human-auditable policy

[A1] rl_agent.py (reward, ε-decay) · [A3] run_simulation Module 3 (convergence 713/800)

Routing Network Topology Graph — BFS + RL Routing

[A2] topology.py (241-node graph, BFS) · [A1] rl_agent.py (RL routing on the graph)

Security Quantum Key Distribution

BB84 Protocol

1. Alice sends random polarised photons (H/V, D/A) [13]
2. Bob measures in random bases
3. Public reconciliation — keep matching bases
4. Estimate QBER from a sampled subset
5. Privacy amplification (universal hashing)
6. QBER < 11% → SECURE key established [15]

Post-Quantum & Deployment

• ML-KEM (FIPS 203) — key encapsulation [16]
• ML-DSA (FIPS 204) — digital signatures [17]
• Hybrid classical + quantum key exchange

Phase	Protocol	Key rate
1 — LEO	BB84	1–10 kbps
2 — GEO	BB84 + E91	10–100 kbps
3 — Mars	E91 + repeaters	1+ kbps

[13] Bennett-Brassard 1984 · [14] Ekert 1991 · [15] Shor-Preskill 2000 (QBER<11%) · [16][17] NIST FIPS 203/204 · [A5] qkd.py

Orbital Contact Windows & Doppler Compensation

Mars Parameters

Semi-major axis	1.524 AU
Synodic period	779.94 days
Distance range	54.6M – 401M km
Areostationary alt.	17,032 km
Doppler (Ka, max)	±2.1 MHz @ 24 km/s

Contact Windows

Window	Avail.	Duration	Rate
Optimal	99%	8–12 hrs	100–200 Mbps
Good	95%	6–8 hrs	20–100 Mbps
Fair	85%	4–6 hrs	5–20 Mbps
Blackout	0%	2–4 wks	—

[3] JPL Horizons (synodic 779.94 d) · [A2] topology.py · doppler.py (±2.1 MHz Ka @ 24 km/s)

LO-e Radiation-Hardened Computing

Effects & Mitigation

Effect	Mitigation
SEU (bit flip)	SECDED ECC
MBU (adjacent bits)	Bit interleaving
SEL (latch-up)	Current limit + power-cycle
TID (cumulative dose)	Rad-hard parts (RAD750) [18]

Defense-in-Depth

1. TMR — 3 replicas + majority voter
2. SECDED (39,32) ECC — correct 1, detect 2
3. Memory scrubbing before 2nd upset
4. FDIR + watchdog → isolate → reload
5. SAFE-MODE after recovery budget exhausted

37,159

raw upsets

200×

protection factor

3,334×

TMR gain

2,127×

RAD750 TID margin

DEMONSTRATED — [A6] radiation.py, run_simulation Module 6 · [18] BAE RAD750 · [19] ESA LEON3FT

LO-f Mission-Critical Data Prioritization

Priority Tiers

Tier	Class	Examples
P0	Emergency	Health, collision, faults
P1	Mission-critical	Command ACKs, time-sensitive sci
P2	High	Routine telemetry, scheduled sci
P4	Bulk	Housekeeping, files, SW images

Three Levers

• Compression: Rice ≈3× [7], wavelet ≈10× [8]
• Deadline-aware preemptive QoS [A7] — strict priority, earliest deadline
• BPv7 fragmentation [9] — send what fits, defer rest, link never idle

5/6

items delivered

100%

link utilisation (target)

41%

SW update fragmented

P0

emergency preempts

[A7] prioritization.py · [7] CCSDS 121.0-B-3 · [8] CCSDS 122.0-B-2 · [9] RFC 9171 (fragmentation)

Impact Performance vs Current Systems

Metric	Current (MRO) [1]	AETHERIX [A4]	Note
Downlink rate	0.5–6 Mbps	2–200 Mbps	10–100× capability
Availability	60–75%	>95%	design target
Routing	Static (CGR)	RL-adaptive	autonomous [A1]
Security	AES-256	QKD + PQC	future-proof [16][17]
Scalability	5–10 assets	241 nodes	10–100× [A2]
Conjunction	Blackout	L4/L5 relay (target)	+50–70% (target) [A8]

Demonstrated results and design targets are labelled separately — the 10–100× figure is a physics-derived capability from the link budget, not a measured throughput.

[1] NASA MRO · [A4] link_budget.py · [A2] topology.py · [A8] Module 4 · targets clearly labelled

Tools Implementation & Standards

33

Python modules

643

tests (37 files)

15

interactive demos

5

routing policies

Core Modules

• Link Budget · RL Agent · QKD
• Bundle Protocol · Topology · Simulation
• Radiation · Prioritization · Doppler

All physics is real — no mocked data.

Standards Compliance

RFC 9171	Bundle Protocol v7
RFC 4838 / 5326 / 7242	DTN / LTP / TCPCL
RFC 9172	BPSec (security)
CCSDS 734.2 / 734.3	Bundle Protocol / SABR
CCSDS 141.0-B-1	Optical comms physical

[9][10][11][12] IETF RFCs · [2][5] CCSDS · [A1]–[A8] AETHERIX modules (github.com/matx104/AETHERIX)

Design Trade-off Analysis — Why These Choices

Decision	Choice	Rationale vs alternative
Optical vs RF	Hybrid 1550 nm + Ka fallback [4]	10–100× throughput [A4]; RF survives clouds & corona
Routing	Custom Q-learning, not ION-DTN CGR [A1]	Adapts to live state; CGR re-plans on stale schedule
RL model	Q-tables now, DQN later	Every value auditable; trains in seconds
Reward weights	α=1.0, δ=10.0, ε-decay 0.995 [A1]	Drop penalty 10× delivery — forbids bundle loss

Each choice sacrifices peak theoretical performance for auditability and reproducibility — exactly what a defence and a research artefact require. DSOC heritage: NASA flew optical + RF side-by-side on Psyche [4].

[A1] rl_agent.py / training.py · [A4] link_budget.py · [4] NASA DSOC (Psyche) · [A2] topology.py

Resilience Failure & Recovery

Autonomous solar-conjunction link-blackout survival

Path	Band	Quality	Status	Reward [A8]
Direct Mars → Earth	1550 nm optical	0.05	CLOSED	−1.438
Mars → ES-L4 → Earth	Ka-band RF	0.65	OPEN	−0.201
Mars → ES-L5 → Earth	Ka-band RF	0.60	OPEN	−0.201

How AETHERIX Recovers (automatically)

1. Detect — optical Q-value collapses (q<0.3)
2. Re-route — agent picks highest-Q: ES-L4 (Ka-band)
3. Prioritise — P0 forwarded on best link; P4 stored & deferred

Why Lagrange Relays

• ES-L4 / ES-L5 sit 60° ahead/behind Earth
• Keep line-of-sight to Mars around the solar limb
• Direct: 0% at conjunction → via L4/L5: 50–70% retained [A8]

Throughput drops (optical → RF) but no mission-critical data is lost. Run live: python run_simulation.py --module 4

[A8] run_simulation Module 4 · [A1] rl_agent.py (0.3 threshold) · [3] JPL Horizons (Lagrange) · [A2] topology.py

Conclusion AETHERIX Delivers

Problem Solved

TCP/IP can't work at interplanetary distances.

What We Built

BPv7 · RL Routing · QKD · Optical/RF Hybrid.

10–100×

target speedup ♦

>95%

availability (target)

RL

adaptive routing [A1]

QKD

quantum security [13][15]

Demonstrated: 241-node topology, RL routing, QKD, 200× radiation reduction. ♦ Targets are labelled as targets, never measured results.

[1] NASA · [3] JPL Horizons · [9][12] IETF · [13][15] QKD · [A1][A2][A4][A5][A8] AETHERIX

References Industry & Scientific [1]–[20]

Three-layer attribution: [N] industry · [AN] this project · every entry links to its source

[1] NASA JPL, Mars Relay Network User's Guide, 2024.

[2] CCSDS, Bundle Protocol Spec., 734.2-B-1, 2015.

[3] JPL Horizons, Earth–Mars Ephemeris, 2025.

[4] NASA, Deep Space Optical Comm. (DSOC), Psyche, 2024.

[5] CCSDS, Optical Comm. Physical Layer, 141.0-B-1, 2019.

[6] CCSDS, TM Synchronization & Channel Coding, 131.0-B-5, 2023.

[7] CCSDS, Lossless Data Compression, 121.0-B-3, 2020.

[8] CCSDS, Image Data Compression, 122.0-B-2, 2017.

[9] IETF, Bundle Protocol Version 7, RFC 9171, 2022.

[10] IETF, Licklider Transmission Protocol, RFC 5326, 2008.

[11] IETF, DTN TCP Convergence Layer, RFC 7242, 2014.

[12] IETF, DTN Architecture, RFC 4838, 2007.

[13] Bennett & Brassard, Quantum Cryptography, 1984.

[14] Ekert, Bell's-Theorem Cryptography, PRL 67, 1991.

[15] Shor & Preskill, Security of BB84, PRL 85, 2000.

[16] NIST, Module-Lattice KEM, FIPS 203, 2024.

[17] NIST, Module-Lattice Signature, FIPS 204, 2024.

[18] BAE Systems, RAD750 Rad-Hard PowerPC, 2024.

[19] ESA/Gaisler, LEON3FT Processor, 2023.

[20] IETF, RFC 4838 §3.1, 2007 (DTN challenges).

Full URLs: github.com/matx104/AETHERIX · next: project sources [A1]–[A8]

References Project Sources [A1]–[A8]

M. A. Tariq, AETHERIX, github.com/matx104/AETHERIX (2025) · each links to the source file

Design & Code

• [A1] RL Routing Agent — src/routing/rl_agent.py
• [A2] Network Topology — src/orbital/topology.py
• [A3] End-to-End Simulation — run_simulation.py Module 3
• [A4] Optical Link Budget — src/infrastructure/link_budget.py

Security, Resilience & QoS

• [A5] QKD Protocols — src/security/qkd.py
• [A6] Radiation Hardening — src/computing/radiation.py
• [A7] Data Prioritization — src/routing/prioritization.py
• [A8] Failure & Recovery — run_simulation.py Module 4

Three-layer attribution: Layer A industry baseline [1]–[20] · Layer B design decision [A1]–[A8] · Layer C demonstrated result (e.g. radiation 200× [A6]). Design targets are labelled as such, never as measured results.

Reproducible: python run_simulation.py · 643 tests: python -m pytest tests/ -q · matx104.github.io/AETHERIX

THANK YOU

Questions?

10–100×

target speedup ♦

>95%

availability (target)

RL

adaptive routing

QKD

quantum security

Muhammad Abdullah Tariq

matx104.github.io/AETHERIX · github.com/matx104/AETHERIX