Full-Stack Compute Enablement for a Shared Intelligent Robotics Ecosystem
An end-cloud collaborative full-stack technology foundation, partnering across the value chain to accelerate commercial deployment
Full-Stack Compute Enablement for a Shared Intelligent Robotics Ecosystem
An end-cloud collaborative full-stack technology foundation, partnering across the value chain to accelerate commercial deployment
Industry Solutions
Providing precisely matched collaboration models for roles across the full value chain
Full-Stack Enablement for Robot Manufacturers
Provides a full-stack technology suite spanning chips, systems, and cloud services to reduce foundational R&D investment and significantly shorten time to market.
Integrated Delivery for Healthcare and Senior Care Institutions
Pairs terminal robots with a cloud operations platform to deliver 24-hour safety companionship, improve service quality, and ease staffing pressure.
Compliance Co-Building for the Healthcare Ecosystem
Built on an end-to-end medical-grade compliance architecture, co-creates closed-loop health data scenarios and connects with the broader healthcare ecosystem.
Industry Solutions
Providing precisely matched collaboration models for roles across the full value chain
Full-Stack Enablement for Robot Manufacturers
Provides a full-stack technology suite spanning chips, systems, and cloud services to reduce foundational R&D investment and significantly shorten time to market.
Integrated Delivery for Healthcare and Senior Care Institutions
Pairs terminal robots with a cloud operations platform to deliver 24-hour safety companionship, improve service quality, and ease staffing pressure.
Compliance Co-Building for the Healthcare Ecosystem
Built on an end-to-end medical-grade compliance architecture, co-creates closed-loop health data scenarios and connects with the broader healthcare ecosystem.
Edge-cloud Collaborative Full-stack Architecture
Four-layer technology stack collaboration creates a complete system for device-side perception, system control, cloud cognition, and operations hub.
Edge-cloud Collaborative Full-stack Architecture
Four-layer technology stack collaboration creates a complete system for device-side perception, system control, cloud cognition, and operations hub.
Customized Edge AI SoC Chip
A customized edge AI system-on-chip developed specifically for health and eldercare robots. It follows a software-hardware integrated co-design philosophy, is deeply optimized for unstructured home scenarios and human-machine physical interaction, and comprehensively rebuilds device-side inference capabilities.
Technical Feature Description
Customized Edge AI SoC Chip
A customized edge AI system-on-chip developed specifically for health and eldercare robots. It follows a software-hardware integrated co-design philosophy, is deeply optimized for unstructured home scenarios and human-machine physical interaction, and comprehensively rebuilds device-side inference capabilities.
Technical Feature Description
Millisecond-level Multimodal Heterogeneous Fusion Perception
Uses a highly optimized heterogeneous computing architecture with a built-in physical interaction hardware acceleration engine. At the silicon layer, it parallel-schedules RGB-D depth vision, millimeter-wave spatial radar, and flexible force-sensing array data to complete offline 3D spatial modeling and dynamic obstacle avoidance.
Hardware-level Trusted Execution Environment (TEE)
Builds an encrypted security sandbox at the physical layer. All home vision, voice, and vital-sign data is desensitized and feature-extracted locally; core data remains local throughout the process, minimizing compliance risk.
Localized Quantized Large-model Inference Engine
Applies dedicated architectural optimization for large models. With high tensor compute efficiency, the device side smoothly runs quantized LLM/VLM models to support natural language understanding, emotion recognition, and local companionship dialogue.
Ultra-low-latency Force Feedback and Compliant Control
Reserves an independent ultra-low-latency control data path for real-time processing of high-sensitivity flexible tactile and torque sensor feedback. A kilohertz (kHz) refresh rate drives the robotic arm to deliver human-like compliant force control and reduce rigid collision injury.
Milliwatt-level Always-on Energy-efficiency Architecture
Uses advanced power management. In standby mode, only milliwatt-level power is used to maintain wake-word listening and fall detection. After threshold triggering, full compute wakes rapidly to provide 24-hour all-weather safety redundancy.
Robot Real-time Control System Platform
The core hub connecting device-side computing power with upper-layer applications. Through instruction-set-level software-hardware collaboration, it converts AI semantic instructions into precise robotic motion control, completing the last kilometer from compute to behavior.
Core Technical Points
Instruction set level software-hardware co-design
Low-level real-time control algorithms and custom compilers are deeply integrated with the self-developed chip at the instruction-set level, reducing scheduling latency from traditional operating systems and efficiently converting compute power into closed-loop electromechanical control.
Multimodal Sensor Fusion Scheduling
Uniformly schedules multi-source sensor data such as vision, radar, force sensing, and vital signs, optimizing data transmission paths to achieve low-latency, highly synchronized environmental perception and status feedback.
Edge-cloud Collaborative Intelligent Routing
A built-in dynamic task scheduling engine intelligently determines task priority and complexity. Basic real-time tasks are processed locally, while complex cognitive tasks collaborate with the cloud, balancing latency, cost, and results.
Robot Real-time Control System Platform
The core hub connecting device-side computing power with upper-layer applications. Through instruction-set-level software-hardware collaboration, it converts AI semantic instructions into precise robotic motion control, completing the last kilometer from compute to behavior.
Core Technical Points
Instruction set level software-hardware co-design
Low-level real-time control algorithms and custom compilers are deeply integrated with the self-developed chip at the instruction-set level, reducing scheduling latency from traditional operating systems and efficiently converting compute power into closed-loop electromechanical control.
Multimodal Sensor Fusion Scheduling
Uniformly schedules multi-source sensor data such as vision, radar, force sensing, and vital signs, optimizing data transmission paths to achieve low-latency, highly synchronized environmental perception and status feedback.
Edge-cloud Collaborative Intelligent Routing
A built-in dynamic task scheduling engine intelligently determines task priority and complexity. Basic real-time tasks are processed locally, while complex cognitive tasks collaborate with the cloud, balancing latency, cost, and results.
AI Service SaaS Cloud Brain
An enterprise-grade SaaS platform built specifically for health and eldercare companion robots. It provides advanced cognitive capabilities and full-lifecycle operational enablement for medical institutions, health and eldercare centers, and robot manufacturers, creating a cloud hub for intelligent companionship and care.
AI Service SaaS Cloud Brain
An enterprise-grade SaaS platform built specifically for health and eldercare companion robots. It provides advanced cognitive capabilities and full-lifecycle operational enablement for medical institutions, health and eldercare centers, and robot manufacturers, creating a cloud hub for intelligent companionship and care.
