统一数据流与低复制传输 Unified dataflow and low-copy transfer
同一数据对象可沿统一数据流提供给多个节点，减少不必要的重复搬运。 A data object can serve multiple nodes through one dataflow, reducing unnecessary duplication.
Technology / Dora OS
让数据、算法与设备在同一运行底座上协同 One runtime for data, models and devices
Dora 是面向机器人和具身智能系统的数据流与运行底座，连接传感器、算法模型、任务节点和执行设备，为数据通信、设备管理、任务编排与运行监控提供统一支撑。 Dora is a dataflow and runtime foundation for robotics and embodied AI. It connects sensors, models, task nodes, and actuators while supporting communication, device management, orchestration, and runtime monitoring.
Runtime foundation
从设备连接到运行监控 From device connection to runtime monitoring
Dora把通信、设备、算法节点、任务和运行状态放到同一套运行机制中，使机器人系统中的不同组件能够交换数据并协同工作。 Dora brings communication, devices, algorithm nodes, tasks, and runtime state into one mechanism so robotics components can exchange data and work together.
01
数据通信
Data communication
连接不同节点的数据输入与输出
Connect inputs and outputs across nodes
02
设备管理
Device management
组织传感器、执行器和设备接口
Organize sensors, actuators, and interfaces
03
节点管理
Node management
管理算法与功能节点的运行关系
Manage algorithm and function nodes
04
任务编排
Task orchestration
连接任务步骤和执行顺序
Connect task steps and execution order
05
运行监控
Runtime monitoring
观察节点、数据流和运行状态
Observe nodes, dataflow, and runtime state
Unified dataflow
让数据沿明确的路径运行 Move data through explicit paths
Dora以节点和数据流组织系统，把设备驱动、算法模型、任务逻辑与运行监控连接起来。高频小数据和大体量数据可以根据各自特点进入相应处理路径。 Dora organizes systems as nodes and dataflows, linking device drivers, models, task logic, and monitoring. High-frequency small messages and large data can follow paths suited to their characteristics.
设备驱动
Drivers
算法节点
Algorithms
任务节点
Tasks
Dora Dataflow
连接数据区与布局元数据
Connect data regions and layout metadata
记录
Record
回放
Replay
监控
Monitor
System-level optimization
围绕数据传输、时间与计算位置优化 Optimize transfer, time, and compute placement
Dora的底层设计关注机器人系统中反复出现的三个问题：跨节点数据传输、多模态时间对齐，以及端侧与边缘计算协同。 Dora focuses on three recurring robotics challenges: data transfer across nodes, multimodal time alignment, and coordination between endpoints and edge compute.
全模态时间域记录 Full-modal time-domain recording
将视觉、雷达、音频和机器人状态等多路数据对齐到统一时间轴，用于记录与回放。 Align vision, LiDAR, audio, and robot state to one timeline for recording and replay.
端
End
边
Edge
任务
Task
03 / Coordination
端点—边缘异构协同 Endpoint-edge coordination
连接端侧设备、边缘计算单元与任务节点，让不同计算位置共同完成任务。 Connect endpoint devices, edge compute, and task nodes across heterogeneous compute locations.
Record and replay
把一次运行保留下来，用于回放、分析和训练 Preserve each run for replay, analysis, and training
多路设备数据经过采集和时间对齐后进入统一记录，用于问题复现、运行分析、模型评测与训练数据连接。 Device streams are captured and time-aligned into a unified record for issue reproduction, runtime analysis, model evaluation, and training connections.
01
设备驱动
Drivers
02
多流采集
Capture
03
时间对齐
Align
04
录制归档
Archive
05
回放分析
Replay
06
训练与复现
Train & reproduce
Development and deployment
连接不同开发环境与计算平台 Connect development environments and compute platforms
Dora面向机器人研发中的多语言、跨平台和分布式运行需求，为算法节点、设备接口和任务流程提供一致的数据流组织方式。 Dora addresses multilingual, cross-platform, and distributed robotics development with a consistent way to organize algorithm nodes, device interfaces, and task flows.
语言接口
Language interfaces
Rust · Python · C/C++
运行环境
Runtime environments
Linux · Windows
计算位置
Compute placement
端侧 · 边缘 · 分布式节点
Endpoint · Edge · Distributed nodes
验证环境
Validation environments
Isaac Sim · OasisBot
Dora
Open Runtime
开发者
Developers
机器人与设备
Robots & devices
算法与模型
Algorithms & models
仿真工具
Simulation tools
科研与产业
Research & industry
Open ecosystem
围绕开放运行底座连接工具与社区 Connect tools and communities around an open runtime
Dora以开放方式连接开发者、机器人设备、算法模型、仿真工具以及科研和产业实践，使不同组件能够进入同一数据流与运行环境。 Dora openly connects developers, robotic devices, models, simulation tools, and research and industry practice within one dataflow and runtime environment.
Technology relationship