自动检测
英语
请输入需要翻译的内容 ✨ Project Overview U-Care Yuka is a robot system designed for elderly care, family safety, and intelligent services. The project uses a cloud-edge-end collaboration architecture, where real-time interactions, model inference, and physical actions are distributed across different computing nodes:
Emotional Interaction: ASR, TTS, LLM Agents, digital humans, and knowledge base queries. Home Security: Fire, smoke, and fall detection, with cloud-based preliminary filtering and edge-level verification. Embodied Operations: Mobile chassis navigation, dual-arm collaboration, and visual language-based tasks. Cross-Device Collaboration: Communication between Raspberry Pi, ECS, Ascend, and Jetson via HTTP, WebSocket, or Socket. Health Analysis: Health data accessed through MCP with user authorization; the repository does not contain any real personal health records.
This repository represents a source-only version suitable for GitHub release. Model weights, training data, runtime logs, virtual environments, front-end dependencies, and large demonstration media are not included in version control.
🧭 System Architecture
View the system architecture diagram 🧩 Core Features Feature | Description | Main Module Emotional Voice Interaction | Streaming speech recognition, synthesis, agent dialogue, and digital human control | opentalking/ Cloud AI Agent | LLM, MCP, management console, and service orchestration | yuka_agent/ Fire and Fall Detection | MindSpore/MindYOLO inference and edge verification | ai_server/ Edge Perception | Cameras, microphones, alerts, and cross-device scheduling | edge_pi/ Mobile and Dual-Arm Operations | ROS2 navigation, task orchestration, and LeRobot dual-arm control | jetson_robot/ Chassis Real-Time Control | STM32F407VE, FreeRTOS, motor and sensor control | stm32/
📁 Project Structure ├── ai_server/ # MindSpore/Ascend fire and fall detection ├── edge_pi/ # Raspberry Pi multi-modal interaction terminal ├── jetson_robot/ # ROS2 mobile chassis, dual-arms, and LeRobot ├── stm32/ # STM32F407VE chassis firmware ├── opentalking/ # Digital humans, voice, RAG, and web front-end ├── yuka_agent/ # LLM deployment, MCP, ECS services, and management console ├── assets/ # Lightweight demonstration images required for README ├── LICENSE # Apache License 2.0 └── THIRD_PARTY_NOTICES.md
There are significant differences in the environment for each module; please refer to the corresponding documentation:
AI Security Service Description Raspberry Pi Edge Description Jetson Mobile Robot Description Dual-Arm LeRobot Description STM32 Chassis Firmware Description Yuka Agent LLM Deployment Description Yuka Agent Server Description OpenTalking English Description
🚀 Quick Start
- Get the source code git clone cd U-Care-Yuka-Emotional-Interactive-Robot-Based-on-lerobot
- Choose the nodes to deploy This project is not a standalone application; all modules do not need to be started simultaneously. It’s recommended to select the appropriate node based on the specific scenario:
Scenario | Recommended Entry Point Experience digital humans and voice interaction | opentalking/ Deploy ECS Agent and management console | yuka_agent/yuka_agent_server/ Deploy Ascend LLM / voice cloning | yuka_agent/llm_deploy/ Run fire and fall detection | ai_server/ Configure Raspberry Pi interaction interface | edge_pi/ Configure Jetson navigation and dual-arms | jetson_robot/ Compile STM32F407VE chassis firmware | stm32/USER/*.uvprojx
- Prepare external resources Models, datasets, and runtime resources should be downloaded or generated according to the documentation for each module. It’s recommended to store them in the designated directory; the root .gitignore file prevents accidental commits.
Common external resources include: LLM, ASR, TTS, QuickTalk, VLA, and vision detection model weights; Robot collected data, training sets, and checkpoints; Digital human character materials, facial caches, and generated videos; User private data such as Apple Health; .env files, API Keys, device tokens, and server keys.
🔐 Security and Privacy Do not submit .env files, private keys, API Keys, device tokens, or actual service addresses. Health data must be explicitly authorized by the user and stored only locally or in controlled services. It’s recommended to run GitHub Secret Scanning, Gitleaks, or similar tools before uploading. If any security issues are found, refer to SECURITY.md; do not disclose keys or vulnerability details in public issues.
🤝 Contribution Feel free to submit issues and pull requests. Before doing so, read CONTRIBUTING.md and ensure that:
No models, datasets, caches, build artifacts, or personal data are included; Original copyright and licenses are preserved when adding third-party code; New features require updates to the corresponding documentation and configuration examples.
🙏 Thank You Thank you to MindSpore, MindYOLO, LeRobot, OpenTalking, xiaozhi-esp32-server, FreeRTOS, CMSIS, and STM32 software ecosystem, as well as partners who provided computing power, hardware, and engineering support for this project.
📄 Open Source License The original code in this project uses Apache License 2.0 by default.
The repository contains or adapts several third-party projects, which are still governed by their respective licenses. For detailed information on paths, sources, and licenses, see THIRD_PARTY_NOTICES.md. Some STM32 component licenses are proprietary but allow certain re-distributions, and thus are not automatically covered by Apache-2.0. Model weights, datasets, fonts, sounds, and character materials may also have separate licenses; please check the relevant terms before use.