Simulations

Overview

HASH is designed as an end-to-end solution for quickly creating massive, rich, real-time simulations.

Currently, as a developer you can use HASH's standalone simulation tools to build rich, detailed, agent-based simulations of factory floors, business processes, public policy decisions, or any other type of complex system you can imagine.

Our simulation docs are split into:

• Creating a simulation
• Running simulations

Future Plans

While HASH's simulation engine (hEngine) and IDE (hCore) currently exist as standalone projects, their functionality will be integrated to the core HASH application.

This will allow us to offer a number of key improvements over simulations currently runnable using hEngine and hCore.

Better, easier-to-create simulations

With Generative AI, we're able to deliver a much-improved agent-based modeling experience:

• More easily define simulation behaviors: domain experts will be able to define agents' behaviors using natural language, convertible to Python/JavaScript/Rust code executable in HASH.
• Ready-to-go simulation agents: simulation variables, agents and their attributes can be be instantiated at the point of need, based on semantic entities already in your web, without the need for any data pre-processing.
• Runtime dynamism: unhandled scenarios that occur within simulations can be automatically handeld by allowing generative AI to program an agents' next steps.
• Generative 3D assets: unique 3D-generated representations of entities can be generated based on its attributes, bringing visualizations to life (custom meshes, textures, and in the future realistic physics and character animations with ActionGPT-like or dual DRL-generative approaches).

More powerful simulations

• Human-in-the-loop simulation: unhandled scenarios that occur within simulations can be identified, and execution paused for domain expert input.
• Automatic continuous calibration: models' initial states are automatically inferred based on the latest live data in a HASH web.
• Cloud execution platform: simulation runs and experiments will be executable as part of flows. Previously available as hCloud.

More useful outputs

• Better views: existing, separated views (Geospatial/3D and the Step Explorer/Charts) will be integrated into more powerful, combined views -- allowing 3D assets to be overlaid on maps, and providing access to more featureful charts and graphs.
• API-accessible outputs: entities in webs can be easily accessed via the HASH API.

Getting Started

As a new user, the best way to learn more about agent-based modeling is to read up on the basics, and try your hand at the simulation tutorials. You can also browse our public registry of pre-built HASH models to find ones with interesting dynamics that you can freely use and adapt for your own needs.

If you've used other simulation modeling software before, you might find that models built in HASH are a little different to those you're used to seeing.

HASH leverages composable agents, behaviors and types to make it fast and easy to create realistic, well-calibrated simulations.

Simulation IDE

hCore is a browser-based integrated development environment (IDE) for building agent-based models and running simulations. It provides a temporary graphical user interface for working with simulations until such time as they are natively supported within HASH itself. hCore will ultimately be deprecated, and is no longer officially supported by HASH.

Simulation Engine

hEngine is HASH's simulation engine. In the future we plan to integrate this directly into the main HASH application. For now, an experimental version of it is available to use independently. Interested parties with experience developing distributed systems in Rust, and an interest in simulation modeling, are invited to get involved.

Getting Help

We're not yet offering general support for developing simulations in HASH. However, if you need help with a simulation, or have questions about using HASH for simulations, feel free to contact us directly at any time.