Source: https://broxme.com/developers

Developers 

# Robotics developer resources.

Learn how Studio represents robot projects, runs programs in simulation, and captures evidence for debugging.

Start with the engineering guide and a browser example. SDK contracts and installation guidance will follow verified releases.

[Read the engineering guide](https://broxme.com/robotics-development-guide) [Explore the Robotics SDK](https://broxme.com/robotics-sdk)

Developer Tools 

## A software foundation for robotics experiments.

Studio engineering work covers programming, project models, simulation, and diagnostics. Release verification varies by workflow.

### Robotics Programming

Studio’s supported visual, Python, and C++ workflows share a representation for robot behavior.

### Project Models and APIs

Studio’s internal project model keeps robot definitions and program state together. The public Core interfaces are in development.

[Explore Broxme Core](https://broxme.com/core)

### Simulation and Testing

Use deterministic runs and replay to investigate behavior in the engineering build.

### Hardware Interfaces

Engineering work covers transport interfaces and firmware generation. Physical release testing is pending.

### ROS 2 Interoperability

Studio’s engineering work includes a bridge and a browser-tested setup workflow. Live ROS 2 graph and physical-robot verification remain open.

### Diagnostics and Reliability

The engineering build includes runtime inspection, local recovery, and flight recording workflows.

### Extensible Workflows

Studio’s engineering work on extensions informs development of the public Robotics SDK.

[Explore the SDK](https://broxme.com/robotics-sdk)

Developer Documentation 

## Know where the interfaces meet.

The engineering guide explains the current project, execution, and recording boundaries. Public installation steps, API reference, and a supported-hardware matrix await release verification.

### Project and model formats

Studio’s .nrobot project remains the source of project state. Engineering import and export for URDF and SDF are bounded and can lose unsupported semantics.

[Read the project workflow](https://broxme.com/robotics-development-guide#studio-architecture)

### Recording and replay

Engineering tests cover recording, replay, and Studio’s MCAP profile. A recording captures run evidence separately from project state.

### Before testing hardware

Physical and native release validation varies by board, transport, platform, and ROS 2 configuration.

[Read the Studio status](https://broxme.com/studio#development-status)

Development Approach 

## Make experiments repeatable.

We use clear interfaces, test before deployment, and prefer established standards where they fit the workflow.

### Specify the boundaries

Record the software version, project format, and inputs that define the test.

### Separate kinds of evidence

Source tests, simulation, and native or physical-device runs each answer different questions.

### Document what carries over

Record compatibility limits alongside examples so developers can see which assumptions apply.

Robot learning workflow 

## Collect, train, and evaluate a policy.

Use one task with a measurable outcome to trace the workflow. Record the robot configuration, observation streams, action representation, and evaluation scenarios alongside each policy version.

### Record complete episodes

Keep synchronized camera frames, robot state, actions, task labels, and episode boundaries. Hold out entire episodes or scenarios for evaluation so adjacent frames from the same run stay together. LeRobotDataset v3 provides one published representation for this kind of data.

[Read the LeRobot dataset format](https://huggingface.co/docs/lerobot/lerobot-dataset-v3)

### Train and compare policies

Imitation learning trains a policy from demonstrations. A recent NVIDIA Isaac GR00T workflow shows how teams can collect teleoperated demonstrations, fine-tune a policy, and evaluate it in Isaac Lab Arena. Keep a simple baseline and compare both policies on the same scenarios.

[Read NVIDIA’s GR00T workflow](https://developer.nvidia.com/blog/develop-humanoid-robot-policies-end-to-end-with-nvidia-isaac-gr00t/)

### Review results by scenario

Report completed trials, timeouts, contacts, and interventions with the total trial count. Keep traces for the failure cases and look for patterns across starting poses, lighting, and obstacle placement.

[Read the evaluation workflow](https://broxme.com/robotics-development-guide#evaluation-workflow)

## Start with a repeatable example.

Observe how the scripted rover responds to a barrier. Then see how the Studio runtime handles a virtual run.

[Run the browser example](https://broxme.com/robotics-development-guide#rover-example)