AMD is a leader in high-performance and adaptive computing, committed to providing high-quality products and services to help customers solve their biggest challenges. Our technolo…
Jingfeng Microcontroller’s robot open platform based on AMD Ryzen processorUsing AMD ROCm for reinforcement learning in robot simulation environmentLingbao CASBOT02 robotXinbu Embodied Intelligent Development Platform HB03-X100
AMD is a leader in high-performance and adaptive computing, committed to providing high-quality products and services to help customers solve their biggest challenges. Our technologies drive the data center, embedded systems, gaming and PC markets into the future.
Founded in 1969 in Silicon Valley with just a few dozen employees, AMD has been innovating and leading the way in semiconductor products ever since. Today, AMD has grown into a modern, global company setting new standards for modern computing with advanced technologies and breakthrough industry innovations.
02
MAIN PRODUCTS
Key Products
06 ITEMS
01
Jingfeng Microcontroller’s robot open platform based on AMD Ryzen processor
The product is based on the bipedal robot brain platform built on AMD Embedded Ryzen™ 8840U, which fully utilizes the heterogeneous computing capabilities of CPU, iGPU and NPU to realize a number of core functions such as robot lidar mapping and autonomous navigation, environment perception and autonomous movement, voice dialogue interaction, and gesture command recognition control, allowing the robot to respond naturally to user operations. The platform provides complete robot secondary development documents and open source code. Users can quickly optimize algorithms, scene adaptation and application expansion based on existing functions.
02
Using AMD ROCm for reinforcement learning in robot simulation environment
This demo shows the strategic deployment of different reinforcement learning algorithms for various tasks of bipedal robots, quadrupedal robot dogs, and robotic arms. It can realize complex actions such as walking, operating and even climbing on walls and flipping in complex collision environments.
03
Lingbao CASBOT02 robot
CASBOT02 is the second-generation full-size bipedal humanoid robot of Zhongke Huiling (Lingbao CASBOT). It focuses on "lightweight, high interaction, and multi-scenario adaptation" and is widely used in commercial tours, entertainment and performing arts, scientific research and education, etc. In this demo, CASBOT02 cooperates with AMD Ryzen AI X100 computing platform and localized large model inference engine to achieve low-latency multi-modal interaction on the device side.
04
Xinbu Embodied Intelligent Development Platform HB03-X100
The platform main control uses AMD Ryzen AI
05
Huayan seven-axis humanoid arm
Huayan Robotics, together with AMD, can apply the heterogeneous computing power of Zynq-7000 to the optimization of robot control systems, achieve microsecond-level response and multi-axis collaboration, have human-like tactile perception and force interaction capabilities, and have full-joint high-precision force control accuracy ≤0.02N·m, end-connection The touch force accuracy reaches 0.1N, and it can sense small external forces and respond compliantly like a human hand; with seven degrees of freedom and a crisscross wrist joint design, combined with advanced motion control algorithm capabilities and system-level AI capabilities, it can replicate the flexible movements of the human arm such as obstacle avoidance, detouring, and arm bending, allowing it to operate calmly in small spaces.
06
Submillimeter-level robot full-link simulation training solution
AMD joins hands with Pedai Technology to launch an embodied intelligent simulation training solution based on the ROCm platform, with "reality" running through the entire link. Core capabilities include: sub-millimeter level three-dimensional reconstruction to replicate reality at a 1:1 level, a true physical optical mechanics engine that accurately restores light and shadow collisions, and supports the free import of digital models to quickly expand scenes. The solution covers a closed-loop process from real data cold start, high-fidelity scene restoration, synthetic data expansion and post-reinforcement learning deployment verification.
