Robot I/O Modes

robot_io controls how the policy reads robot state and sends low-level commands. It is separate from:

• inference_backend, which controls ONNX / TensorRT policy inference.
• motion_backend, which controls the reference motion source.

For sim2sim with sim_env/base_sim.py, keep the policy on --robot-io zmq. For sim2real deployment, choose one of the three modes below.

Quick Choice

Mode	Processes	Use when
--robot-io inline --robot g1	policy only	Preferred G1 path when the policy runs on a machine with unitree_interface; avoids the extra ZMQ bridge hop.
--robot-io zmq + scripts/g1/real_bridge.py	policy + Python DDS bridge	Use the G1 split-process bridge based on unitree_sdk2py.
--robot-io zmq + scripts/g1/real_bridge_cpp.py	policy + unitree_interface bridge	Keep the ZMQ contract while using the unitree_interface robot binding.

tip

If hardware tests show violent joint shaking, or high-dynamic / high-speed motions look much worse than expected, suspect unstable ZMQ I/O latency first. Try --robot-io inline before retuning the policy or gains.

Inline

Inline mode creates the G1 RobotIO backend inside BasePolicy. The policy runtime reads normalized RobotState values and sends commands through the backend without starting a bridge process.

uv run sim2real/rl_policy/tracking.py \
  --robot-io inline \
  --policy-config checkpoints/mimic-lite/roa/policy.yaml

Use this path first for real deployment when latency jitter matters. Add --robot-interface <robot_network_interface> only when the robot network interface is not the default eth0.

ZMQ With scripts/g1/real_bridge.py

This mode keeps the policy and robot bridge in separate processes. The bridge uses unitree_sdk2py, publishes low_state over ZMQ, and applies low_cmd from ZMQ to the robot.

Terminal 1:

uv run scripts/g1/real_bridge.py

Terminal 2:

uv run sim2real/rl_policy/tracking.py \
  --policy-config checkpoints/mimic-lite/roa/policy.yaml

Add --interface <robot_network_interface> to the bridge command only when the robot network interface is not the default eth0.

ZMQ With scripts/g1/real_bridge_cpp.py

This mode keeps the same ZMQ contract as scripts/g1/real_bridge.py, but the bridge uses unitree_interface for robot I/O.

Terminal 1:

uv run scripts/g1/real_bridge_cpp.py

Terminal 2:

uv run sim2real/rl_policy/tracking.py \
  --policy-config checkpoints/mimic-lite/roa/policy.yaml

Add --interface <robot_network_interface> to the bridge command only when the robot network interface is not the default eth0.

Rule Of Thumb

Use inline for the normal real-robot deploy path. Use ZMQ bridge mode when you need process isolation, want to debug the policy and robot bridge separately, or are running sim2sim with base_sim.py.