Skip to main content

RDKS100_LNX_SDK_V4.0.2

The RDK S100 Linux SDK includes the following components:

  • RDK Linux: Includes the Linux Kernel, Ubuntu desktop environment, hardware drivers, etc.
  • OpenExplorer (OE): A toolkit for model compilation and optimization, algorithm repository, and application development SDK
  • OpenExplorer-LLM (OE-LLM): Toolchain tailored for large language models
  • TogetheROS.B: A ROS2-based robotics development kit providing ROS2 packages with hardware acceleration support and pre-trained models
  • D-Navigation: Image flashing tool

RDK Linux​

OS​

  • Linux Kernel 6.1.112-rt43
  • Ubuntu 22.04 Desktop-based D-Robotics customized system
    • Supports RDK-specific commands

Hardware Interface​

  • SPI controller support
  • I2C controller support
  • UART controller support
  • WiFi + Bluetooth (UART) support

Multimedia API​

  • HBN support: Serdes, MIPI, CIM, ISP, YNR, PYM, GDC, STITCH
  • V4L2: Same as above except STITCH
  • Sample programs:
    • sample_cim: CIM streaming example
    • sample_isp: RAW image processing example
    • sample_gdc: Distortion correction example
    • sample_pym: Image scaling example
    • sample_gpu: GPU example
    • sample_isp_feedback: ISP feedback injection example
    • Video pipeline examples:
      • mipi → cim → isp → ynr → pym → vpu → Save as H264/H265
      • mipi → cim → isp → ynr → pym → Offline GDC
      • serdes(raw) → mipi → cim → isp → ynr → pym → Offline GDC → vpu
      • serdes(yuv) → mipi → cim → pym → Offline GDC → vpu
      • USB camera example
    • sunrise_camera: Real-time streaming example via browser or VLC

BPU​

  • Supports OpenExplorer 3.2.0

Graphics​

  • HDMI supports 2.5K@60fps display

GPU​

  • Supports:
    • EGL 1.0~1.5
    • OpenGL ES 1.1~3.2
    • Vulkan 1.0~1.3
    • OpenCL 1.0~3.0

Security​

  • OP-TEE version: 3.19.0

MCU SDK​

OS​

FreeRTOS v10.0.1

Sample Programs​

  • Supports CAN pass-through communication to Linux
  • UART transmission example
  • SPI transmission example
  • PWM waveform generation example
  • Ethernet MAC layer transmission example
  • ADC channel voltage reading example
  • GPIO control example

OpenExplorer​

OE Development Toolkit​

Provides complete examples for model quantization and deployment, supports adaptation and optimization of multiple mainstream models, and helps users quickly build efficient inference pipelines. Includes the following:

  • Local development environment installer
  • Post-Training Quantization (PTQ)-based model quantization examples
  • Quantization-Aware Training (QAT)-based model quantization examples
  • Model deployment and performance evaluation examples

OE User Manual​

Provides usage instructions and configuration guidance for all OpenExplorer components, covering installation, toolchain invocation, model adaptation workflows, etc., suitable for both beginners and engineering developers.

CPU Docker Environment​

Provides an OpenExplorer Docker image for CPU, integrating necessary dependencies. It must be used together with the OE Development Toolkit for model compilation, quantization, and basic deployment testing.

GPU Docker Environment​

Provides a Docker runtime environment for GPU, supporting GPU-based model optimization, inference testing, and tuning workflows. It must also be used in conjunction with the OE Development Toolkit.

OpenExplorer-LLM​

info

Not released in this version. Please contact FAE for access.

tip

The commercial version offers more comprehensive feature support, deeper hardware capability exposure, and exclusive customization. To ensure compliance and secure delivery, we will grant access to the commercial version through the following process:

Commercial Version Access Process:

  1. Fill out a questionnaire: Submit basic information about your organization and use case.
  2. Sign an NDA: We will contact you based on your submission, and both parties will sign a Non-Disclosure Agreement upon mutual confirmation.
  3. Content release: After the NDA is signed, we will provide access to the commercial version materials via a private channel.

If you wish to obtain the commercial version, please fill out the questionnaire below. We will contact you within 3–5 business days:

Questionnaire link: https://horizonrobotics.feishu.cn/share/base/form/shrcnJQBMIkRm6K79rjXR0hr0Fg

TogetheROS.B​

hobot_sensor​

Adaptation components for commonly used robot sensors, supporting USB and MIPI camera integration, lowering development barriers and shortening integration cycles, allowing developers to focus on core business logic.

hobot_dnn​

A lightweight inference framework designed specifically for on-device robot inference scenarios, simplifying model deployment and fully leveraging BPU computing power to lower the technical barrier for deploying intelligent algorithms on edge devices.

hobot_codec​

A software-hardware integrated video codec module that significantly reduces CPU resource usage and enhances parallel processing capabilities. Supports efficient conversion between MJPEG, H264, H265 and image formats such as BGR8, RGB8, and NV12.

hobot_cv​

An accelerated module for common computer vision operators, supporting efficient resize of NV12 format images and transcoding between NV12 and BGR24 formats, improving algorithm execution efficiency and reducing system load.

hobot_render​

A visualization rendering module supporting multiple display methods for real-time presentation of algorithm results, facilitating debugging and demonstrations. Compatible with mainstream tools such as web browsers, RViz2, and Foxglove Studio.

Boxs Algorithm Repository​

A pre-trained intelligent algorithm library provided for robot manufacturers and ecosystem developers, enabling rapid integration and deployment of various vision and perception models on D-Robotics RDK systems. Includes the following model categories:

  • Object Detection: YOLO series, Faster R-CNN, EfficientDet, etc.
  • Image Classification: MobileNetV2
  • Semantic Segmentation: MobileNet-UNet, YOLOv8-Seg
  • Application Algorithms:
    • Human detection and pose tracking: Yolo-Pose
    • Speech perception processing: SenseVoice
    • Environmental perception and mapping: BEV, CenterPoint
    • Depth estimation and stereo vision: Stereo depth estimation
    • Multimodal perception: InternVL (Vision-Language Model)
    • SLAM support: DFMatch-based feature extraction and matching

Apps Application Examples​

End-to-end application examples integrating image input, perception processing, and decision-making strategies, demonstrating a complete perception-to-intelligence pipeline and helping developers quickly build demo projects. Examples include:

  • Multimodal Smart Box
  • Vision-and-Language Interactive System (ASR + VLM/LLM + TTS)

D-Navigation​

  • Supports Windows, Linux, and Mac
  • Supports eMMC/UFS image flashing

Known Issues​

warning
  • RDK Linux
    • RTC YSN8130E is not yet supported in software
    • BMI088 IMU on the MCU daughterboard is not yet supported in software
    • Some Ubuntu applications crash under GPU acceleration due to page size mismatch
  • MCU SDK
    • The current MCU system has some stability issues
    • BMI088 IMU on the MCU daughterboard is not supported
    • Lockstep unlock is currently unsupported
Last updated on