RDK S100 Series
Human-Inspired Brain Architecture
Driving the Co-Evolution of Intelligence in
Embodied Robots
Most Intelligent Brain + Most Efficient Small Brain
Perception + Decision + Action Tri-Unity
Body Intelligence Has Never Been Smarter and Simpler
Humanoid-Inspired Dual-Brain Architecture
Integrating CPU + BPU + MCU for super-heterogeneous computing and high-efficiency coordination
6x ARM Cortex-A78AE
CPU
100/137K DMIPS
Supports real-time kernels for
low-latency, intelligent scheduling
Featuring the next-generation
BPU Nash
80/128 TOPS
1
Supports 160+ ONNX standard operators
optimized for CNNs and Transformers
delivering faster inference with lower
power consumption
Powerful
Cerebrum
High-Performance
Cerebellum
4x Arm Cortex-R52+
MCU
2
6K+ DMIPS
Enables high-frame-rate, low-latency
joint control for precise real-time motion execution
Large & Small Models Efficient Collaboration
Dynamic fusion and seamless switching from perception to execution
OmniVision — Large model with multi-category support
Supports vision and LiDAR detection, LLMs, VLMs, and a wide range of popular AI models.
Vision Model Inference Like Lightning
Multiple motion control models supported — MCU and BPU collaborate to lower CPU load and manage more tasks.
Reduced
CPU
Usage
80%
Seamless Model Collaboration
Perception, Decision & Control working together efficiently for complex, dynamic real-world scenarios.
Hand-Eye Unlimited Expansion
Meets a variety of needs for research, testing, and prototype development
All-in-One Development
Seamless HW-SW collaboration and end-to-cloud integration enable effortless development
Official Support
Official support for popular AI models, easy to deploy
Edge–cloud collaboration
End-to-end closed-loop data system for collection, training, and deployment
Hardware Acceleration
Official hardware-accelerated ROS2 packages for improved performance and efficiency
Easy Optimization
Comprehensive model adaptation and optimization solutions for faster algorithm deployment
Flash-Link: A Game-Changing Development Experience
Connect, develop, and advance with a single Type-C cable, unlocking the full potential of RDK S100.
Flash Connect development flow chart, showing the full process experience from board connection, development and usage to advanced development
Application Rapid Deployment
200+ open-source algorithms and application examples for faster multi-scenario deployment
Extensive Accessories
Camera Expansion Board
MCU Port Expansion Board
Full-Metal Protective Case
GMSL Camera
MIPI Camera
More Accessories Coming Soon
Parameter

RDK S100RDK S100P

AI Performance
80 TOPS128 TOPS
CPU
6x Arm Cortex-A78AE@1.5GHz6x Arm Cortex-A78AE@2.0GHz
MCU
4x Arm Cortex-R52+@1.2GHz
GPU
1x Arm Mali-G78AE@100 GFLOPS
Memory
12GB LPDDR5, 96-bit24GB LPDDR5, 96-bit
Storage
64GB eMMC
M.2 Key M (PCIe 3.0 x1) * 1
Display
HDMI 1.4 up to 2K@60FPS
Network
Giqabit Ethernet(RJ45 Connector)*2
M.2 Key E(PCle 3.0 x1,UART)Connector for Wi-Fi&Bluetooth
Ports
USB 2.0 (Type-C) * 1 for flash & debug
USB 3.0 * 4
Expansion Header
JTAG Header * 1
Automatic Header * 1
40-pin Header for Main GPlO * 1
l6-pin Header for MCU GPlO * 1
Camera Expansion Header * 1
MCU Port Expansion Header *1
Power
12-20V DC
Mechanical
120x121x51mm (Case Included)
Notes:
1. Equivalent computing power with INT8 quantization.
2. Includes one pair of DCLS cores running in lockstep and two cores supporting Split-Lock mode, allowing flexible choice between lockstep or independent operation based on application needs.
3. BPU handles motion-control model inference, while MCU manages motion command interpolation and high-frequency communication, reducing CPU load.

RDK S100 Series

Human-Inspired Brain Architecture
Driving the Co-Evolution of Intelligence in Embodied Robots

RDK S100 算控一体机器人开发套件宣传图,展示产品集成优势

Humanoid-Inspired Dual-Brain Architecture

Integrating CPU + BPU + MCU for super-heterogeneous computing and high-efficiency coordination

Powerful Cerebrum

High-Performance
Cerebellum

RDK S100 Heterogeneous Architecture Diagram, integrating CPU, BPU, and MCU to achieve efficient collaborative computing
6x ARM Cortex-A78AECPU

100/137K DMIPS

Supports real-time kernels for low-latency,
intelligent scheduling
Featuring the next-generationBPU Nash

80/128 TOPS1

Supports 160+ ONNX standard operators, delivering
faster inference with lower power consumption
4x Arm Cortex-R52+ MCU2

6K+ DMIPS

Enables high-frame-rate, low-latency joint
control for precise real-time motion execution

Large & Small Models Efficient Collaboration

Dynamic fusion and seamless switching from perception to execution

Schematic diagram of RDK S100 model performance, showing the vision/point cloud processing capabilities of large models and the detection performance of multiple models

OmniVision — Large model with multi-category support

Supports vision and LiDAR detection, LLMs, VLMs, and a wide range of
popular AI models.

Vision Model Inference Like Lightning

Multiple motion control models supported — MCU and BPU collaborate to lower CPU load and manage more tasks.

Reduced
CPU
Usage

80%

Seamless Model Collaboration

Perception, Decision & Control working together efficiently for complex, dynamic real-world scenarios.
RDK S100 Model Collaboration Diagram, showing the perception - decision - control collaboration ability

Hand-Eye Unlimited Expansion

Meets a variety of needs for research, testing, and prototype development

RDK S100 Interface Expansion Diagram, marking 12 types of interfaces, including diverse interfaces such as the 40PIN classic interface and commonly used MCU interfaces, supporting the diverse development needs of robots.

All-in-One Development

Seamless HW-SW collaboration and end-to-cloud integration
enable effortless development

  • Official Support

    Official support for popular AI models, easy to deploy
  • Edge–cloud collaboration

    End-to-end closed-loop data system for collection, training, and deployment
  • Hardware Acceleration

    Official hardware-accelerated ROS2 packages for improved performance and efficiency.
  • Easy Optimization

    Comprehensive model adaptation and optimization solutions for faster algorithm deployment

Flash-Link: A Game-Changing Development Experience

Connect, develop, and advance with a single Type-C cable, unlocking the full potential of RDK S100.

Flash Connect development flow chart, showing the full process experience from board connection, development and usage to advanced development

Application Rapid Deployment

200+ open-source algorithms and application examples for faster multi-scenario deployment

Extensive Accessories

More Accessories Coming Soon

  • The exterior view of the RDK S100 Camera Expansion Board showcases an industrial-grade board structure, with the fiber optic interface and connector layout clearly presented.

    Camera Expansion Board

  • Appearance diagram of RDK S100 MCU interface expansion board, showing the layout of interfaces such as network ports and green terminals, to facilitate hardware function expansion.

    MCU Port Expansion Board

  • Appearance diagram of RDK S100 all-metal protective housing, with a compact design paired with metal material, enhancing equipment protection and industrial texture

    Full-Metal Protective Case

  • Schematic diagram of the RDK S100 GMSL camera appearance, showing its lens, body, and hardware form adapted to GMSL technology

    GMSL Camera

  • RDK S100 MIPI Camera Appearance Diagram, showing its board-level module design, MIPI flexible cable interface, and compatibility with embedded vision solutions

    MIPI Camera

ParameterRDK S100RDK S100P

  • AI Performance

    80 TOPS128 TOPS

  • CPU

    6x Arm Cortex-A78AE@1.5GHz6x Arm Cortex-A78AE@2.0GHz

  • MCU

    4x Arm Cortex-R52+@1.2GHz
  • GPU

    1x Arm Mali-G78AE@100 GFLOPS
  • Memory

    12GB LPDDR5, 96-bit24GB LPDDR5, 96-bit

  • Storage

    64GB eMMC
    M.2 Key M (PCIe 3.0 x1) * 1
  • Display

    HDMI 1.4 up to 2K@60FPS
  • Network

    Gigabit Ethernet (RJ45 Connector) * 2
    M.2 Key E (PCIe 3.0 x1, UART) Connector for Wi-Fi & Bluetooth
  • Ports

    USB 2.0 (Type-C) * 1 for flash & debug
    USB 3.0 * 4
  • Expansion Header

    JTAG Header * 1
    Automatic Header * 1
    40-pin Header for Main GPIO * 1
    16-pin Header for MCU GPIO * 1
    Camera Expansion Header * 1
    MCU Port Expansion Header * 1
  • Power

    12-20V DC
  • Mechanical

    120x121x51mm (Case Included)

Notes:

1. Equivalent computing power with INT8 quantization.
2. Includes one pair of DCLS cores running in lockstep and two cores supporting Split-Lock mode, allowing flexible choice between lockstep or independent operation based on application needs.
3. BPU handles motion-control model inference, while MCU manages motion command interpolation and high-frequency communication, reducing CPU load.