Bonsai simulation partners

Simulations support reinforcement learning with the Bonsai Machine Teaching platform with interactive environments for iterative training. In addition to supporting common coding languages like Python, Bonsai works with software vendors to support commercial simulation platforms.

Commercial simulation partnership falls into two categories:

  1. Fully integrated: Fully integrated software vendors provide the full suite of Bonsai simulation functionality and appear as a supported simulator upload type in the Bonsai UI.
  2. Works with Bonsai: Software platforms that work with Bonsai meet basic requirements for validating that simulations built on their platform will train successfully with Bonsai.

Important

Users who choose to build their simulation with supported partners may need to obtain additional licenses or a commercial agreement from the vendor to train AI at scale.

Simulator platform support

Fully integrated partners

Platform Vendor Resources
AnyLogic AnyLogic microsoft/bonsai-anylogic
Gym OpenAI microsoft/bonsai-gym
Simulink MathWorks microsoft/bonsai-simulink
Python Custom microsoft/microsoft-bonsai-api
VP Link Wood capesoftware/bonsai-vplink

Fully integrated software platforms provide the following benefits:

  • Bonsai can automatically run and scale training with your simulation.
  • Brain training is validated to work with publicly available tooling and samples.
  • Easily connect unmanaged simulations with native connection tools.
  • Easily create managed simulation packages by uploading your simulation.
  • Programmatically generate a brain from managed simulation packages.
  • Public documentation for connecting and using unmanaged simulators build on the platform.
  • Support from Bonsai and the software vendor for integration problems.
  • Billing and commercial agreements for training AI at scale are handled directly through Bonsai.
  • Tools, connectors, and libraries that provide integration logic are publicly available with pre-negotiated billing.

"Works with Bonsai" partners

The following platforms work with Bonsai, but have not met the requirements to qualify for full integration. If you have a simulation model running on one of these platforms, and need additional assistance or resources, please contact us for help.

Platform Vendor Resources
Amesim (FMU) Siemens No public resources
Cosmo Tech Cosmo Tech No public resources
Dynamic Simulation (DYNSIM) Aveva No public resources
FlexSim FlexSim No public resources
Flomaster (FMU) Siemens microsoft/FMU-bonsai-connector
Fluent (Digital Twin Builder) ANSYS microsoft/bonsai-twin-builder
gPROMS Siemens PSE No public resources
Hadean Distributed Cloud Platform Hadean No public resources
MATLAB MathWorks microsoft/bonsai-simulink
MuJoCo MuJoCo No public resources
Petro-SIM KBC No public resources
Robot Operating System (ROS) Open Source microsoft/ROS-bonsai-connector
SimPy Open Source NealAnalyticsLLC/bonsai_simpy_demo
Tecnomatix Plant Simulation Siemens No public resources
Tecnomatix Process Simulate Siemens No public resources

Platform requirements for partners

"Works with Bonsai" software requirements

To qualify for "Works with Bonsai" status, vendors must:

  1. Implement a simulation model for integration verification.
  2. Implement an Inkling file for a sample brain that uses the verification model.
  3. Provide integration support such that Bonsai can start training, step through training iterations, and finish an episode loop with the simulator.
  4. Prove the verification model performs fast enough to be useable with Bonsai (<30 sec/iteration).
  5. Prove the corresponding sample brain in Bonsai can connect to the verification model as an unmanaged simulator and successfully learn the required concept.
  6. Establish a custom licensing agreement for customers.

"Fully integrated" software requirements

To qualify for "Fully integrated" status, vendors must:

  1. Meet all requirements for "Works with Bonsai" platforms.
  2. Update their simulation software to run as a containerized (e.g., Docker) application.
  3. Implement model inspection to generate Bonsai-compatible JSON describing all relevant states and actions in the model.
  4. Provide native integration tools for unmanaged simulators build on the platform.
  5. Agree to a formal payment schedule for scaling and licensing.
  6. Establish a support process to help software customers troubleshoot Bonsai integrations.
  7. Update public sales collateral to highlight use cases with Bonsai.

Connector template

To ensure a consistent experience for customers, simulation software vendors should use the Bonsai connector repo template to structure the customer-facing GitHub repo for their integration connector.