Uranus

Uranus is the world simulator and digital twin engine of the Menlo Stack. It provides rapid scenario testing and validation—simulation where iteration happens before hardware is ever at risk.

Why We Built Uranus

Simulation is where iteration happens. Before any code touches a physical robot, we need to validate behavior, stress-test edge cases, and confirm readiness. Uranus compresses the feedback loop from weeks to hours.

What Uranus Provides

Uranus serves three purposes:

1. Scenario Generation

Uranus produces high-fidelity, customer-relevant scenarios to stress-test agents before deployment. These scenarios are derived from real-world operational data and edge cases, ensuring agents are prepared for deployment conditions—not just happy-path scenarios.

2. Pre-Deployment Validation

Uranus enables regression testing and safe exploration in a risk-free environment. Before an agent touches a physical robot, we validate that it behaves correctly across hundreds of scenarios. This catches failures early, when they’re cheap to fix.

3. Hardware-in-the-Loop Testing

Uranus supports virtual commissioning to reduce real-world testing time. By simulating the physical robot alongside software agents, we validate the full system—agent, motor control, and environment—without requiring physical hardware.

Digital Twin Architecture

Uranus maintains digital twins of:

• Physical environments: Warehouses, factories, and other deployment sites
• Asimov robots: High-fidelity models of deployed hardware
• Customer scenarios: Reconstructions of real operational situations

These digital twins are continuously updated with data from real-world deployments, ensuring simulation fidelity improves over time.

Integration

Uranus connects:

• Cyclotron: Shares scenario data and receives trained policies for validation
• Agent Platform: Provides validation environment before physical deployment
• Data Engine: Consumes real-world data to improve scenario fidelity
• Asimov: Receives validated policies and contributes operational data back

The Feedback Loop

Uranus compresses the development cycle:

1. Agent is defined in agent framework
2. Agent is tested against Uranus scenarios
3. Failures are identified and remediated
4. Cyclotron trains new motor skills if needed
5. Agent re-validated in Uranus
6. Only then does physical deployment occur

This loop is measured in hours, not weeks.