physicalai-runtime-running-policy-on-robot

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Running a Policy on a Robot

在机器人上运行策略

PolicyRuntime

(

src/physicalai/runtime/runtime.py

) owns the control loop;

InferenceModel

owns policy math. Execution strategy lives under

src/physicalai/runtime/execution.py

and

src/physicalai/runtime/rtc_execution.py

. CLI:

physicalai run

src/physicalai/cli/run.py

instantiates from YAML via jsonargparse.

PolicyRuntime

(

src/physicalai/runtime/runtime.py

) 负责控制循环；

InferenceModel

负责策略运算。执行策略定义在

src/physicalai/runtime/execution.py

和

src/physicalai/runtime/rtc_execution.py

中。CLI命令：

src/physicalai/cli/run.py

中的

physicalai run

通过jsonargparse从YAML文件实例化相关组件。

Workflow

工作流程

Choose API vs config: Python for notebooks/tests; YAML +
```
physicalai run
```
for reproducible deployment.
- Done when: entry point matches the user's task.

Python minimal loop (see

docs/how-to/runtime/run-policy-on-robot.md

python

from physicalai.runtime import PolicyRuntime, SyncExecution
from physicalai.inference import InferenceModel
from physicalai.robot import SO101
from physicalai.capture import UVCCamera

runtime = PolicyRuntime(
    fps=30,
    robot=SO101(port="/dev/ttyACM0"),
    model=InferenceModel("./exports/act_policy"),
    cameras={"wrist": UVCCamera(device="/dev/video0", width=640, height=480)},
    execution=SyncExecution(),
)
with runtime:
    runtime.run(duration_s=60)

Done when: components connect and the loop runs in a test or dry-run with fakes.

YAML config — nest

class_path

init_args

for robot, model, cameras, execution; run:

bash

physicalai run --config runtime.yaml --run.duration_s=60

Execution mode — pick sync vs RTC per
```
docs/how-to/runtime/use-execution-modes.md
```
; do not build ad-hoc timing around
```
InferenceModel.select_action
```
when
```
PolicyRuntime
```
should own the queue.
Callbacks — register via runtime callback APIs (
```
docs/how-to/runtime/add-runtime-callbacks.md
```
) for telemetry/latency, not inside inference adapters.

选择API或配置文件：在笔记本/测试场景使用Python代码；在可复现的部署场景使用YAML配置 +
```
physicalai run
```
命令。
- 完成标志：入口点与用户任务匹配。

Python最小化循环（参考

docs/how-to/runtime/run-policy-on-robot.md

）：

python

from physicalai.runtime import PolicyRuntime, SyncExecution
from physicalai.inference import InferenceModel
from physicalai.robot import SO101
from physicalai.capture import UVCCamera

runtime = PolicyRuntime(
    fps=30,
    robot=SO101(port="/dev/ttyACM0"),
    model=InferenceModel("./exports/act_policy"),
    cameras={"wrist": UVCCamera(device="/dev/video0", width=640, height=480)},
    execution=SyncExecution(),
)
with runtime:
    runtime.run(duration_s=60)

完成标志：组件成功连接，且循环在测试环境或使用模拟设备的试运行中正常运行。

YAML配置 —— 为机器人、模型、摄像头、执行器嵌套
```
class_path
```
/
```
init_args
```
；运行命令：
bash
```
physicalai run --config runtime.yaml --run.duration_s=60
```
执行模式 —— 根据
```
docs/how-to/runtime/use-execution-modes.md
```
选择同步模式或RTC模式；当
```
PolicyRuntime
```
应管理队列时，不要围绕
```
InferenceModel.select_action
```
构建临时计时逻辑。
回调函数 —— 通过运行时回调API（参考
```
docs/how-to/runtime/add-runtime-callbacks.md
```
）注册遥测/延迟相关回调，不要在推理适配器内部实现。

Validation loop

验证循环

bash

uv run pytest tests/unit/runtime/ -q

Use fake robots/cameras from runtime tests when hardware is unavailable.

bash

uv run pytest tests/unit/runtime/ -q

当硬件不可用时，使用运行时测试中的模拟机器人/摄像头。

Required checks

必要检查

```
fps
```
and camera read rates are consistent.
Action dimensions match robot
```
send_action
```
expectations.
Config
```
class_path
```
targets are importable without training packages.
Document breaking changes to runtime config schema in
```
docs/reference/config-schema.md
```
.

```
fps
```
与摄像头读取速率保持一致。
动作维度符合机器人
```
send_action
```
的预期。
配置中的
```
class_path
```
目标无需依赖训练包即可导入。
在
```
docs/reference/config-schema.md
```
中记录运行时配置架构的重大变更。

References

参考资料

docs/how-to/runtime/run-policy-on-robot.md

docs/how-to/config/write-runtime-config.md

docs/how-to/runtime/run-policy-on-robot.md

docs/how-to/config/write-runtime-config.md