On June 11, D-Robotics launched the RDK S100, described as the industry’s first all-in-one system-on-chip (SoC) development kit for robots with a human-inspired architecture. The system is designed to support collaboration between large and small artificial intelligence models, aiming to close the loop between perception, decision-making, and control. This represents an essential step in advancing embodied intelligence by enabling tighter coordination between a robot’s cognitive and motor functions.
The RDK S100 features a broad array of peripheral interfaces and offers full-stack support across hardware-software integration and edge-to-cloud workflows. It is intended to help developers more efficiently build embodied robots and deploy them across diverse, real-world environments.
Currently, embodied intelligence is evolving along two primary paths: end-to-end architectures and hierarchical decision models. The latter is generally considered easier to interpret and implement. However, their adoption hinges on the availability of flexible, power-efficient SoC platforms that can handle varied computational tasks. The RDK S100 is designed to meet these requirements.
By integrating a CPU, a brain processing unit (BPU), and a microcontroller unit (MCU) on a single chip, the RDK S100 aims to reduce hardware costs while enhancing overall system performance. This design enables the system to allocate high-level processing to decision-making functions while maintaining responsive motor control, supporting coordinated, real-time operations across the robot’s subsystems.
The hybrid architecture allows dynamic switching and fusion between large and small AI models. This setup supports a wide range of complex tasks—large models, such as vision-language models (VLMs) or large language models (LLMs), are used for perception and decision-making, while smaller models manage fast, precise control of motor functions. The integration of the MCU and BPU is said to reduce CPU workload by up to 80%, improving system responsiveness.
For example, consider a scenario where a robot is tasked with retrieving a cup. Upon receiving a voice command, the CPU and BPU activate their visual-semantic detection functions, analyzing visual input and contextual data while accounting for the robot’s current state. Based on this information, the BPU generates a motion plan, and the MCU executes it in real time, enabling the robot to complete the task with precision.
The kit includes a variety of peripheral interfaces to support research, testing, and prototyping. To complement the hardware, D-Robotics also provides a development environment that includes AI modeling tools, hardware acceleration packages, model optimization kits, and a library of more than 200 open-source algorithms and application demos.
In a media interview, Hu Chunxu, vice president overseeing the developer ecosystem at D-Robotics, categorized the robotics landscape into three broad segments:
- Traditional robots, such as industrial systems and robotic vacuum cleaners, which operate in a mature, competitive market.
- Emerging robot formats tailored for niche uses, including pool cleaners, lawn mowers, ball collectors, and companion robots. These focus on solving specific problems through intelligent features.
- Embodied robots, such as humanoid and large quadruped systems, which must perform complex perception and decision-making tasks, requiring significantly more computing power.
Acknowledging the different computational demands across these categories, Hu said that D-Robotics plans to release additional boards with varied performance specifications alongside the RDK S100 to support a wider range of use cases.
According to the company, it has formed partnerships with more than 20 companies working on embodied intelligence and is conducting trials of the RDK S100 with over 50 clients. The kit is priced at RMB 2,799 (USD 391.9) and is compatible with a range of robotic systems, including those requiring bird’s eye view (BEV) or multi-stream video detection.
Looking ahead, D-Robotics plans to introduce more high-performance offerings in the first quarter of 2026 to further support the development of advanced embodied robotics.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This article was written by Huang Nan for 36Kr.