Inspiration
Europe’s critical services increasingly depend on interconnected terrestrial and space infrastructure. A disruption rarely affects only one component: flooding can damage fibre routes, cyberattacks can compromise gateways, interference can degrade satellites, and congestion can reduce network capacity.
We built OrbitResilience AI to explore a central question:
When critical connectivity fails, how can operators restore it quickly without compromising sovereignty, security, or human accountability?
What it does
OrbitResilience AI is an interactive mission-assurance platform for sovereign hybrid satellite and 5G networks.
The application models a digital twin containing:
- Private 5G infrastructure
- Terrestrial fibre links
- European gateways
- LEO and MEO satellites
- Capacity, latency, interference, and cyber-risk signals
- Data-sovereignty and mission policies
During the demo, the network begins in a healthy state. A simulated correlated incident then disrupts terrestrial infrastructure and degrades critical traffic delivery to 41%.
OrbitResilience AI:
- Detects the affected infrastructure.
- Enumerates alternative recovery paths.
- Estimates the modeled risk of each path.
- Evaluates latency, capacity, security, and sovereignty.
- Explains why unsuitable alternatives were rejected.
- Sends the best compliant route through an independent assurance gate.
- Requires explicit human authorization before restoration.
- Generates an auditable decision receipt.
In the demonstration scenario, the selected recovery path is:
Munich 5G → Vienna Gateway → LEO-22 → LEO-17 → Madrid Gateway
The route provides 41 ms latency, 69% available capacity, and 26% modeled risk, while satisfying the required European sovereignty policy.
How we built it
We created the prototype with dependency-free HTML, CSS, and JavaScript, making it fast to load, easy to inspect, and accessible without an account or API key.
The browser-based digital twin represents infrastructure as a graph. When an incident occurs, the decision engine enumerates viable paths through the surviving nodes and links.
Each candidate is evaluated using an interpretable logistic risk model based on synthetic signals including:
- Congestion
- Interference
- Cyber anomalies
- Degraded infrastructure
- Baseline component risk
The risk estimate is combined with latency, available capacity, and mission-policy compliance through a multi-objective scoring process.
Hard requirements—such as data sovereignty and minimum capacity—are implemented separately as policy-as-code constraints. A second assurance layer checks the selected route against mission thresholds before it can be presented for authorization.
This separation is intentional: the optimization engine recommends a route, the assurance gate verifies it, and a human operator remains responsible for the restoration decision.
Challenges we faced
One of our biggest challenges was avoiding a black-box experience where the system simply announced that “AI selected this route.”
For mission-critical infrastructure, operators need to understand:
- Why a route was selected
- Which policies it satisfies
- Why other routes were rejected
- Whether the decision passed independent assurance checks
- Who authorized the final action
We therefore designed the interface to expose candidate routes, rejection reasons, individual metrics, policy results, and the final decision receipt.
Another challenge was balancing competing objectives. The shortest route may have insufficient capacity, cross a prohibited jurisdiction, use degraded infrastructure, or carry unacceptable cyber risk. Resilience cannot be reduced to latency alone.
We also worked to make a technically complex space-and-terrestrial scenario understandable within a short interactive demonstration.
Accomplishments that we are proud of
During the hackathon, we created:
- A functioning hybrid satellite and 5G digital twin
- Graph-based recovery-route enumeration
- A transparent, interpretable risk model
- Multi-objective route ranking
- Deterministic sovereignty controls
- An independent assurance gate
- Human-in-the-loop recovery authorization
- Explainable candidate rejection reasons
- An auditable decision receipt
- A zero-cost, no-login public demonstration
What we learned
We learned that resilient connectivity is not simply about finding the fastest available route.
A trustworthy recovery system must combine:
- Probabilistic risk assessment
- Hard policy constraints
- Mission-specific performance thresholds
- Transparent explanations
- Independent verification
- Accountable human authorization
We also learned that optimization and assurance should remain separate. A system should not be allowed to approve its own recommendation without an additional verification step.
What’s next
Future versions could integrate:
- Public orbital data and live telemetry
- Time-expanded routing for moving satellite constellations
- Advanced anti-jamming and cyber-anomaly models
- Organization-specific sovereignty policy packs
- Incident recording and deterministic replay
- Standards-based satellite and telecom testbeds
- Signed decision receipts for external auditing
- Multi-operator coordination across European infrastructure
OrbitResilience AI currently uses a synthetic topology and simulated telemetry for demonstration purposes. It is an independent hackathon concept and is not affiliated with or endorsed by ESA, the European Union, IRIS², or any satellite operator.
Built With
- 5g
- artificial-intelligence
- css3
- cybersecurity
- digital-twin
- explainable-ai
- graph-algorithms
- graph-optimization
- html5
- human-in-the-loop
- javascript
- leo-satellites
- logistic-regression
- machine-learning
- meo-satellites
- mission-assurance
- network-resilience
- policy-as-code
- render
- satellite-networks
Log in or sign up for Devpost to join the conversation.