Developing with Humanoid Robot Heads: Xiaoling ROS 2 and JINSAN Android

Building a research prototype or a custom AI interaction system? The software interface matters as much as the robot's appearance. Xiaoling and JINSAN A-series heads offer different development paths: Xiaoling's documented platform uses ROS 2, while the JINSAN A-series development materials describe an Android and Java hardware-control application.

This guide helps engineers compare those paths, plan a custom language-model integration, and prepare a focused project brief. Interfaces and deliverables depend on the exact model, hardware revision, software version and agreed project scope.

Xiaoling ROS 2 vs JINSAN Android: what changes for developers?

Choose a platform based on the software environment and the interaction you intend to build, then verify the exact interface package for the selected unit.

The two systems do not imply a shared SDK or interchangeable commands. A ROS 2 integration for JINSAN would need a separate, scoped bridge; the supplied materials do not establish an existing ROS 2 driver for it.

Xiaoling: develop around ROS 2 Humble

The supplied Xiaoling S100 handbook describes a layered system for interaction, motion, peripherals and operations. Its ROS 2 interfaces can help a team divide its project into observable components rather than treating the robot as a single black box.

• Dialog: the handbook lists text input, session start and stop, dialog status and text result interfaces.
• Audio: it describes dialog playback, interruption and speaker-status interfaces.
• Head and sensors: it describes higher-level pose and mouth commands, diagnostics, and selected peripheral events. Availability and permissions depend on the actual configuration.

For example, a research team could begin by observing dialog status and testing one permitted text interaction, then investigate its own model and voice pipeline. Confirm the device's running software and message definitions before relying on any named interface; a published message alone does not prove that the robot processed or spoke it.

Explore the Xiaoling half-body robot and contact us to confirm the development configuration for your project.

JINSAN A-series: build with Android and Java

The supplied A-series materials include Android hardware-control demo projects and separate documentation for 3399 and 3566 versions. They describe combining example files with an application project, calling demo interfaces and building an APK. The 3399 instructions also describe APK signing. Obtain the correct package and signing procedure for the unit rather than assuming the versions are interchangeable.

The example materials describe servo control for features including eyelids, eyes, cheeks, mouth and neck. The permitted movement range, speed and calibration must be confirmed for the specific head. Start with the vendor-approved demo and avoid direct movement commands until the operator and integration engineer have confirmed the limits. Do not put signing files or credentials in a public repository.

From components to a finished face

The supplied production footage shows examples of head components, assembly and face finishing. The exact construction, appearance and interface depend on the selected configuration.

In the supplied workshop footage, an assembled head mechanism is handled beside a development workstation. Mechanical layout and accessible controls depend on the head version; confirm the matching development package before designing around a specific interface.

After the internal components are assembled, the outer face and finishing contribute to the finished presentation. Appearance and manufacturing steps can vary by head, so discuss the desired look and physical compatibility with Warmcore.

Explore the JINSAN AI head or view the A-series head options. Confirm the A-series development package and compatibility before ordering.

Can I connect my own LLM?

Both platforms can be considered for a custom AI interaction project, but the integration work differs. A typical design separates speech input, model inference, speech output and robot control. The model might run on a separate computer or service; the actual deployment depends on the hardware, software and project requirements.

• Xiaoling: the handbook discusses replacing the dialog implementation while retaining its external ROS 2 interfaces. A team must build and test the replacement, audio path, session handling and recovery against the target robot.
• JINSAN A-series: the supplied files demonstrate Android hardware control. Connecting an LLM, speech recognition or speech synthesis would require additional application development and has not been established by these demo files.

These are integration approaches, not a promise that a particular model provider, local model, microphone path or feature is ready out of the box. Keep motion commands within approved limits, and agree on network access and privacy before connecting external services.

A practical first project for research teams

1. Describe one interaction. For example, accept a question and provide a spoken response, with one approved head or facial action if available.
2. Confirm the platform. Record the model, board, software revision, developer environment and interfaces needed for that interaction.
3. Run the matching supplied example. Establish a known baseline with the operator before adding a model or custom application.
4. Add one component at a time. Test text, audio and any approved motion separately. Record what was actually observed on the device.
5. Test interruption and recovery. Agree how the system behaves when an application stops, a service is unavailable or a connection drops.

For more on defining what a supplier provides and what your engineering team builds, see our robot SDK and custom-development guide. Explore the platform features and confirm the configuration relevant to your project.

Developer FAQ

Is every Xiaoling configuration ROS 2 Humble?

The supplied S100 handbook specifies ROS 2 Humble. Confirm the exact software environment and interface package for the model and unit you intend to use.

Is the Xiaoling Python SDK production-ready?

The current repository describes it as an unverified developer preview. Ask about the latest supported release and hardware acceptance before planning a deployment.

Does the JINSAN A-series use the same ROS 2 interface?

The supplied A-series materials describe Android and Java hardware-control examples. They do not establish a shared ROS 2 interface with Xiaoling.

Can an engineer control every movement or sensor?

Not automatically. Access, hardware features, safe limits and feedback depend on the exact model, software and agreed permissions. Confirm them with the integration team.