Joint Wrench Sensor

The joint wrench sensor reports incoming joint reaction wrenches for selected articulation bodies. It exposes force [N] and torque [N·m] buffers separately, with entries ordered by the sensor’s body_names. The default convention is incoming_joint_frame, which expresses each wrench in the child-side joint frame at the child-side joint anchor.

The sensor is configured on an articulation prim and can then be used directly or through manager terms such as body_incoming_wrench(). For example, the Ant environment adds a joint wrench sensor to the scene:

    # robot
    robot = ANT_CFG.replace(prim_path="{ENV_REGEX_NS}/Robot")

    # sensors
    joint_wrench = JointWrenchSensorCfg(prim_path="{ENV_REGEX_NS}/Robot")

The same environment uses SceneEntityCfg to select the reported foot bodies for an observation term:

        feet_body_forces = ObsTerm(
            func=mdp.body_incoming_wrench,
            scale=0.1,
            params={
                "sensor_cfg": SceneEntityCfg(
                    "joint_wrench",
                    body_names=["front_left_foot", "front_right_foot", "left_back_foot", "right_back_foot"],
                )
            },
        )

Direct access to the sensor data follows the usual scene lookup pattern.

joint_wrench = scene["joint_wrench"]
foot_ids, _ = joint_wrench.find_bodies([".*foot"])

force = joint_wrench.data.force.torch[:, foot_ids]
torque = joint_wrench.data.torque.torch[:, foot_ids]
wrench = torch.cat((force, torque), dim=-1)

The resulting wrench tensor has shape (num_envs, num_selected_bodies, 6) and stores the force components followed by the torque components for each selected body.