【Isaac Lab/Isaac Sim】教程4：基于RMPFlow控制6轴机械臂实现真空吸盘抓取

心雨达致

469人浏览 · 2026-05-12 16:17:05

心雨达致 · 2026-05-12 16:17:05 发布

文章目录

1 给6轴机械臂安装真空吸盘
2 调整机器人增益参数
3 导出机器人URDF文件
4 通过Lula Robot Description Editor导出机器人描述参数
5 生成RMPFlow配置文件
6 Standalone 方式通过RMPFlow控制机械臂抓取。
附录：添加VS Code扩展
参考

1 给6轴机械臂安装真空吸盘

教程2：CAD转URDF-基于SolidWorks插件solidworks_urdf_exporter.md

[教程3：基于Surface Gripper制作真空吸盘.md](./教程3：基于Surface Gripper制作真空吸盘.md)

利用教程2和教程3创建的资源组装6轴机器人

打开Isaac Sim
```
cd path/to/isaacsim
./isaac-sim.sh
```
File->New From Stage Template->Empty，创建一个空的Stage
选中World，右键Clear Default Prim
从Content中选择教程2导出的机器人拖入到Viewpoint，并在Stage中选中机器人拖动到Root节点，并右键Set as Default Prim
选中机器人，将Translate属性归零，使机器人位于坐标原点
选中Link6，创建一个Xform，并重命名为ee_Link，并将Transform属性归零为默认值。
选中ee_Link，将其拖到机械手臂root上。
选中ee_Link，添加Rigid Body属性。
选中Link6和ee_Link，创建FixedJoint，重命名为Tool_Joint。

为了方便查看关节，可将Tool_Joint拖动到joints文件夹下
将教程3创建的真空吸盘模型拖入到ee_Link下，并删除Suction_Cup的Rigid Body属性。
展开Suction_Cup，选中suction_cup_joint，Body1点击Add Target，选择Link6
选中旋转关节，点击Play，即可控制机械手臂转动，至此真空吸盘安装成功。

2 调整机器人增益参数

官方教程：Tutorial 11: Tuning Joint Drive Gains

点击Tools->Robotics->Asset Editors->Gain Tunner，打开增益调整器，并点击Play。
调整每个关节的Stiffness和Damping，然后点击SAVE GAINS TO PHYSICS LAYER，然后选择需要测试的关节，调整关节运动范围（默认100%，全部行程），然后点击RUN TEST，等测试结束后，观察波形图，知道波形稳定无毛刺即可。
调整好后点击保存

3 导出机器人URDF文件

Window->Extention，搜索urdf，启用URDF Exporter，并设置为AUTOLOAD
File->Export to URDF，填写文件名，设置Mesh导出路径，点击Export。

在控制台中看到如下输出即为导出成功，可在对应的文件夹下查看导出结果，也可将导出结果上传到Robot Viewer查看.

[8059.134s] [ext: omni.kit.registry.nucleus-0.0.0] startup
self._mesh_dir:  /home/tk/Documents/isaac-proj/i3-V2.0/usd/conf/meshes/
self._mesh_path_prefix:  file://<path to output directory>
self._root:  None
self._visualize_collision_meshes:  True
Root prim not specified. Using the Default Prim on Stage.  /i3_V2_0
Export to URDF successful

请添加图片描述

4 通过Lula Robot Description Editor导出机器人描述参数

官方教程：Tutorial 8: Generate Robot Configuration File

点击Play，然后点击Tools->Robotics->Lula Robot Description Editor，打开Lula机器人描述参数编辑器，Select Articulation选择Stage中的机器人。
将J1-J6设置为Active Joint
Select Link选择不同的Link Prim，然后在Link Sphere Editor中根据实际情况生成碰撞体。
可选择Connect Sphere快速为连杆添加碰撞体。
配置好碰撞体后，导出机器人描述文件。
机器人描述文件导出成功后，点击Tools->Robotics->Lula Test Widget，打开Lula运动学测试工具，然后点击Play。
Select Articulation选择机器人/i3_V2_0，Robot Description YAML加载步骤5导出的机器人yaml描述文件，Robot URDF加载第三节中步骤2导出的urdf文件，然后点击LOAD，Select End Effector Frame选择ee_Link_Suction_Cup_Suction_Cup。

点击Lula Kinematics Solver->FOLLOW TARGET，机器人将运动到红色Box目标位置，且保持末端与Box姿态一致。拖动红色Box，末端将自动跟随。

请添加图片描述

注意

默认的Target位置可能机器人末端无法到达，报下列错误。
2026-05-12T00:16:56Z [63,300,304ms] [Warning] [isaacsim.robot_motion.lula_test_widget.controllers] Failed to compute Inverse Kinematics
所以需要修改默认的Target位置。点击Lula Test Widget 窗口右上角Open Source Code图标，打开扩展源代码，打开test_scenarios.py找到def _create_target(self, position=None, orientation=None)修改target的位置，然后点击保存，重新执行步骤7和8即可。
def _create_target(self, position=None, orientation=None):
 if position is None:
     # target 位置
     position = np.array([0.3, 0, 0.3])
 if orientation is None:
     # target 方向，四元数形式（W，X,Y,Z）
     orientation = np.array([0, -1, 0, 0])
 self._target = VisualCuboid(
     "/World/Target", size=0.05, position=position, orientation=orientation, color=np.array([1.0, 0, 0])
 )

点击Lula Trajectory Generator->CUSTOM TARJECTORY，然后点击FOLLOW TARJECTORY，机器人将移动末端走一个矩形轨迹。

请添加图片描述

注意

默认生成的轨迹点机器人可能无法正常解算出一个合理的动作空间，即报类似下列错误。

[Lula] Task space path conversion reached minimum 's' step size delta' [1e-05] without finding a valid waypoint. The last waypoint conversion was at domain value [0.33], where the domain bounds for the task space path are [0, 1.22361]
[Lula] Conversion from `TaskSpacePathSpec` to `LinearCSpacePath` failed.
2026-05-12T00:40:09Z [64,693,208ms] [Error] [omni.ui.python] AttributeError: 'NoneType' object has no attribute 'get_active_joints'

At:
/home/tk/isaacsim/exts/isaacsim.robot_motion.motion_generation/isaacsim/robot_motion/motion_generation/articulation_trajectory.py(40): __init__
/home/tk/isaacsim/exts/isaacsim.robot_motion.lula_test_widget/isaacsim/robot_motion/lula_test_widget/test_scenarios.py(137): create_trajectory_controller
/home/tk/isaacsim/exts/isaacsim.robot_motion.lula_test_widget/isaacsim/robot_motion/lula_test_widget/extension.py(635): follow_trajectory

所以需要修改默认的轨迹。点击Lula Test Widget 窗口右上角Open Source Code图标，打开扩展源代码，打开test_scenarios.py找到def on_custom_trajectory(self, robot_description_path, urdf_path)修改轨迹点的位置，然后点击保存，重新执行步骤7和9即可。

def on_custom_trajectory(self, robot_description_path, urdf_path):
 self.scenario_reset()
 if self.lula_ik is None:
     return

 self.scenario_name = "Custom Trajectory"

 # 轨迹点方向
 orientation = np.array([0, 1, 0, 0])
 # 轨迹点位置坐标
 rect_path = np.array([[0.3, -0.3, 0.1], [0.3, 0.3, 0.1], [0.3, 0.3, 0.5], [0.3, -0.3, 0.5], [0.3, -0.3, 0.1]])

 self.traj_gen = LulaTaskSpaceTrajectoryGenerator(robot_description_path, urdf_path)

 self._trajectory_base_frame = SingleXFormPrim("/Trajectory", position=np.array([0, 0, 0]))
 for i in range(4):
     frame_prim = self._create_frame_prim(rect_path[i], orientation, f"/Trajectory/Target_{i+1}")
     self._trajectory_targets.append(frame_prim)

一组可行的轨迹点坐标

rect_path = np.array([
                     [0.22, -0.18, 0.28],
                     [0.22,  0.18, 0.28],
                     [0.22,  0.18, 0.20],
                     [0.22, -0.18, 0.20],
                     [0.22, -0.18, 0.28],
             	])

至此机器人描述文件导出成功，并通过测试。

5 生成RMPFlow配置文件

官方文档：

Configuring RMPflow for a New Manipulator

根据官方模板修改为当前机器人配置即可

# Artificially limit the robot joints.  For example:
# A joint with range +-pi would be limited to +-(pi-.01)
joint_limit_buffers: [.01, .01, .01, .01, .01, .01]

# RMPflow has many modifiable parameters, but these serve as a great start.
# Most parameters will not need to be modified
rmp_params:
    cspace_target_rmp:
        metric_scalar: 50.
        position_gain: 100.
        damping_gain: 50.
        robust_position_term_thresh: .5
        inertia: 1.
    cspace_trajectory_rmp:
        p_gain: 100.
        d_gain: 10.
        ff_gain: .25
        weight: 50.
    cspace_affine_rmp:
        final_handover_time_std_dev: .25
        weight: 2000.
    joint_limit_rmp:
        metric_scalar: 1000.
        metric_length_scale: .01
        metric_exploder_eps: 1e-3
        metric_velocity_gate_length_scale: .01
        accel_damper_gain: 200.
        accel_potential_gain: 1.
        accel_potential_exploder_length_scale: .1
        accel_potential_exploder_eps: 1e-2
    joint_velocity_cap_rmp:
        max_velocity: 4.
        velocity_damping_region: 1.5
        damping_gain: 1000.0
        metric_weight: 100.
    target_rmp:
        accel_p_gain: 30.
        accel_d_gain: 85.
        accel_norm_eps: .075
        metric_alpha_length_scale: .05
        min_metric_alpha: .01
        max_metric_scalar: 10000
        min_metric_scalar: 2500
        proximity_metric_boost_scalar: 20.
        proximity_metric_boost_length_scale: .02
        xi_estimator_gate_std_dev: 20000.
        accept_user_weights: false
    axis_target_rmp:
        accel_p_gain: 210.
        accel_d_gain: 60.
        metric_scalar: 10
        proximity_metric_boost_scalar: 3000.
        proximity_metric_boost_length_scale: .08
        xi_estimator_gate_std_dev: 20000.
        accept_user_weights: false
    collision_rmp:
        damping_gain: 50.
        damping_std_dev: .04
        damping_robustness_eps: 1e-2
        damping_velocity_gate_length_scale: .01
        repulsion_gain: 800.
        repulsion_std_dev: .01
        metric_modulation_radius: .5
        metric_scalar: 10000.
        metric_exploder_std_dev: .02
        metric_exploder_eps: .001
    damping_rmp:
        accel_d_gain: 30.
        metric_scalar: 50.
        inertia: 100.

canonical_resolve:
    max_acceleration_norm: 50.
    projection_tolerance: .01
    verbose: false


# body_cylinders are used to promote self-collision avoidance between the robot and its base
# The example below defines the robot base to be a capsule defined by the absolute coordinates pt1 and pt2.
# The semantic name provided for each body_cylinder does not need to be present in the robot URDF.
body_cylinders:
     - name: base
       pt1: [0,0,.333]
       pt2: [0,0,0.]
       radius: .05


# body_collision_controllers defines spheres located at specified frames in the robot URDF
# These spheres will not be allowed to collide with the capsules enumerated under body_cylinders
# By design, most frames in industrial robots are kinematically unable to collide with the robot base.
# It is often only necessary to define body_collision_controllers near the end effector
body_collision_controllers:
     - name: ee_Link
       radius: .05

保存为i3-V2.0_rmpflow_conf.yaml。
点击Tools->Robotics->Lula Test Widget，打开Lula运动学测试工具，然后点击Play.Select Articulation选择机器人/i3_V2_0，Robot Description YAML加载步骤5导出的机器人yaml描述文件，Robot URDF加载第三节中步骤2导出的urdf文件，然后点击LOAD，Select End Effector Frame选择ee_Link_Suction_Cup_Suction_Cup。RmpFlow Config YAML加载保存好的i3-V2.0_rmpflow_conf.yaml配置文件，然后点击FOLLOW TARGET，机器人将自动运动到红色目标中心，并保持末端姿态与目标姿态一致。
至此RMPFlow配置文件测试完成。机器人已经可以通过RMPFlow实现运动学控制了。

6 Standalone 方式通过RMPFlow控制机械臂抓取。

参考官方例程：

standalone_examples/api/isaacsim.robot.manipulators/ur10e

文件结构

i3_v2_0/             					# 项目根目录（建议用下划线，避免横线在 Python 导入时报错）
├── sw_to_urdf/                  		# solidworks导出的urdf文件以及
├── assets/                  			# 存放导出的 .usd 模型和 .urdf
|   ├── meshes/
|   ├── i3-V2.0_with_gripper.usd
│   └── i3-V2.0.urdf
├── configs/                 			# 配置文件
│   ├── i3-V2.0_robot_conf.xrdf      	# 机器人描述文件
│   ├── i3-V2.0_robot_conf.yaml			# 机器人描述文件
│   └── i3-V2.0_robot_rmpflow_conf.yaml	# RMPFlow配置文件
├── controllers/         				# 控制逻辑层
│   ├── __init__.py
│   ├── rpick_palce.py      			# pick and place 控制器
│   └── rmpflow.py 						# RMPFlow 控制器
├── tasks/               				# 高层任务逻辑（场景管理）
│   ├── __init__.py
│   └── pick_place.py					# pick and place 场景
├── .vscode/                 			# VS Code 调试配置
│   └── launch.json          			# 配置 omni.kit.debug.vscode_debugger
└── pick_up_example.py                  # 统一启动入口

项目链接：https://github.com/gitleej/i3_V2_0.git

配置好环境，运行，效果如下。
B站：Isaac Sim 5.0 基于RMPFlow控制六轴机械臂+真空吸盘实现抓取

pick_up_example.py

# -*- coding: utf-8 -*-
# ----------------------------------------------------------------------------------------------------------------------
# pick_up_example.py
# Author:           AILEE
# Create:           2026-04-29 09:43:48
# Last Modified:    2026-05-12 11:31:42
# Last Modified By: AILEE
# Description:      驱动i3机械臂实现拾取和放置的示例脚本
# Copy Right @ 2026 AILEE All Rights Reserved.
# ----------------------------------------------------------------------------------------------------------------------

from isaacsim import SimulationApp
simulation_app = SimulationApp({"headless": False})

from isaacsim.core.api import World
from tasks.pick_palce import PickPlace
from controller.pick_place import PickPlaceController
from omni.isaac.core.utils.math import radians_to_degrees

import numpy as np
np.set_printoptions(precision=3, suppress=True)


my_world = World(stage_units_in_meters=1.0, physics_dt=1 / 200, rendering_dt=20 / 200)
target_position = np.array([-0.3, 0.3, 0.0])
target_position[2] = 0.0515
my_task = PickPlace(
     name="i3_pick_place", 
     target_position=target_position, 
     cube_size=np.array([0.1, 0.1, 0.0515]), 
     cube_initial_position=np.array([0.3, 0.3, 0.026]),
    #  cube_initial_orientation=np.array([0.0, -1.0, 0.0, 0.0])
     )
my_world.add_task(my_task)
my_world.reset()
task_params = my_world.get_task("i3_pick_place").get_params()
r1sc_name = task_params["robot_name"]["value"]
my_r1sc = my_world.scene.get_object(r1sc_name)

# 初始化控制器
my_controller = PickPlaceController(
    name="controller",
    robot_articulation=my_r1sc,
    gripper=my_r1sc.gripper
)

task_params = my_world.get_task("i3_pick_place").get_params()
articulation_controller = my_r1sc.get_articulation_controller()

reset_needed = False
task_completed = False

while simulation_app.is_running():
    my_world.step(render=True)
    if my_world.is_playing():
        if reset_needed:
            my_world.reset()
            reset_needed = False
            my_controller.reset()
            task_completed = False
    
        if my_world.current_time_step_index == 0:
            my_controller.reset()
        
        observations = my_world.get_observations()

        actions = my_controller.forward(
            picking_position=observations[task_params["cube_name"]["value"]]["position"],
            # end_effector_orientation=observations[task_params["cube_name"]["value"]]["orientation"], # 增加姿态输入
            end_effector_orientation=np.array([0, -1, 0, 0]), # 如果不想使用物体姿态，可以设置为固定值
            # placing_position=observations[task_params["cube_name"]["value"]]["target_position"],
            placing_position=target_position,
            current_joint_positions=observations[task_params["robot_name"]["value"]]["joint_positions"],
            # 增加下探距离，确保控制器有“下探”这个动作的空间
            end_effector_offset=np.array([0, 0, 0.024]), 
        )
        if my_controller.is_done() and not task_completed:
                print("done picking and placing")
                task_completed = True
        articulation_controller.apply_action(actions)
        
        end_position = observations[task_params["robot_name"]["value"]]["end_effector_position"]
        end_orientation = observations[task_params["robot_name"]["value"]]["end_effector_orientation"]
        # observations = my_world.get_observations()
        joint_positions = radians_to_degrees( observations[task_params["robot_name"]["value"]]["joint_positions"])
        actions_joint_positions = actions.joint_positions if actions.joint_positions[0] is None else radians_to_degrees(actions.joint_positions)
        # 2. 打印当前状态，排查是否在放置阶段卡住
        current_event = my_controller.get_current_event()
        print("[{}]".format(current_event), "current cube position:", observations[task_params["cube_name"]["value"]]["position"])
        print("[{}]".format(current_event), "current end effector position:", end_position, end_orientation)
        print("[{}]".format(current_event), "actions joint positions:{}".format(actions_joint_positions))
        print("[{}]".format(current_event), "current joint positions:{}".format(joint_positions))

    if my_world.is_stopped():
        reset_needed = True

simulation_app.close()

/task/pick_place.py

# -*- coding: utf-8 -*-
# ----------------------------------------------------------------------------------------------------------------------
# pick_palce.py
# Author:           AILEE
# Create:           2026-04-29 09:50:46
# Last Modified:    2026-05-12 11:14:19
# Last Modified By: AILEE
# Description:      pick place 任务定义
# Copy Right @ 2026 AILEE All Rights Reserved.
# ----------------------------------------------------------------------------------------------------------------------
import os

import isaacsim.core.api.tasks as tasks
import numpy as np

from typing import Optional
from isaacsim.robot.manipulators.manipulators import SingleManipulator
from isaacsim.robot.manipulators.grippers import SurfaceGripper
from isaacsim.core.utils.stage import add_reference_to_stage
from omni.isaac.core.utils.rotations import quat_to_euler_angles
from omni.isaac.core.utils.math import radians_to_degrees

class PickPlace(tasks.PickPlace):
    def __init__(self, 
                 name = "i3_pick_place", 
                 cube_initial_position: Optional[np.ndarray] = None, 
                 cube_initial_orientation: Optional[np.ndarray] = None, 
                 target_position: Optional[np.ndarray] = None, 
                 cube_size: Optional[np.ndarray] = np.array([0.0515, 0.0515, 0.0515]), 
                 offset: Optional[np.ndarray] = None)-> None:
        tasks.PickPlace.__init__(
            self,
            name=name, 
            cube_initial_position=cube_initial_position, 
            cube_initial_orientation=cube_initial_orientation, 
            target_position=target_position, 
            cube_size=cube_size, 
            offset=offset
            )
        return
    
    def set_robot(self):
        current_dir = os.path.dirname(os.path.abspath(__file__))
        assert_path = os.path.abspath(os.path.sep.join([current_dir, "..", "assets/i3-V2.0_with_gripper.usd"]))
        print("assert_path: ", assert_path)
        add_reference_to_stage(
            usd_path=assert_path,
            prim_path="/i3"
        )
        # 定义真空吸盘
        gripper = SurfaceGripper(
            # end_effector_prim_path="/i3/Link6/flange/gripper",
            end_effector_prim_path="/i3/ee_Link",
            surface_gripper_path="/i3/ee_Link/Suction_Cup/SurfaceGripper",
        )
        # 定义机械臂
        manipulator = SingleManipulator(
            prim_path="/i3",
            name="i3_robot",
            # end_effector_prim_path="/i3/Link6/flange/gripper",
            end_effector_prim_path="/i3/ee_Link",
            gripper=gripper,
        )

        return manipulator
    
    def get_observations(self) -> dict:
        """[summary]

        Returns:
            dict: [description]
        """
        joints_state = self._robot.get_joints_state()
        cube_position, cube_orientation = self._cube.get_local_pose()
        end_effector_position, end_effector_orientation = self._robot.end_effector.get_local_pose()
        roll, pitch, yaw = quat_to_euler_angles(end_effector_orientation)
        return {
            self._cube.name: {
                "position": cube_position,
                "orientation": cube_orientation,
                "target_position": self._target_position,
            },
            self._robot.name: {
                "joint_positions": joints_state.positions,
                "end_effector_position": end_effector_position,
                "end_effector_orientation": np.array([radians_to_degrees(roll), radians_to_degrees(pitch), radians_to_degrees(yaw)]),
            },
        }

/controller/pick_place.py

# -*- coding: utf-8 -*-
# ----------------------------------------------------------------------------------------------------------------------
# pick_place.py
# Author:           AILEE
# Create:           2026-04-29 11:27:15
# Last Modified:    2026-05-08 16:57:47
# Last Modified By: AILEE
# Description:      pick place 控制器定义，基于RMPFlow实现
# Copy Right @ 2026 AILEE All Rights Reserved.
# ----------------------------------------------------------------------------------------------------------------------

import isaacsim.robot.manipulators.controllers as manipulators_controllers
from isaacsim.core.prims import SingleArticulation
from isaacsim.robot.manipulators.grippers import SurfaceGripper

from .rmpflow import RMPFlowController

class PickPlaceController(manipulators_controllers.PickPlaceController):
    def __init__(self, 
                 name: str, 
                 gripper: SurfaceGripper, 
                 robot_articulation: SingleArticulation, 
                 events_dt = None) -> None:
        if events_dt is None:
            events_dt = [0.005, 0.002, 1, 0.05, 0.008, 0.005, 0.0008, 0.1, 0.0008, 0.008]
        manipulators_controllers.PickPlaceController.__init__(
            self,
            name=name,
            cspace_controller=RMPFlowController(
                name=name + "_cspace_controller", robot_articulation=robot_articulation
            ),
            gripper=gripper,
            events_dt=events_dt,
            end_effector_initial_height=0.3,
        )
        return

/controller/rmpflow.py

# -*- coding: utf-8 -*-
# ----------------------------------------------------------------------------------------------------------------------
# rmpflow.py
# Author:           AILEE
# Create:           2026-04-29 11:32:13
# Last Modified:    2026-05-12 11:24:23
# Last Modified By: AILEE
# Description:      RMPFlow控制器定义
# Copy Right @ 2026 AILEE All Rights Reserved.
# ----------------------------------------------------------------------------------------------------------------------

import os

import isaacsim.robot_motion.motion_generation as mg
from isaacsim.core.prims import Articulation

class RMPFlowController(mg.MotionPolicyController):
    def __init__(self, 
                 name:str, 
                 robot_articulation: Articulation,
                 physics_dt: float = 1.0 / 60.0) -> None:
        current_dir = os.path.dirname(os.path.abspath(__file__))
        self.rmpflow = mg.lula.motion_policies.RmpFlow(
            robot_description_path=os.path.sep.join([current_dir, "..", "conf/i3-V2.0_robot_conf.yaml"]),
            rmpflow_config_path=os.path.sep.join([current_dir, "..", "conf/i3-V2.0_robot_rmpflow_conf.yaml"]),
            urdf_path=os.path.sep.join([current_dir, "..", "assets/i3-V2.0.urdf"]),
            end_effector_frame_name="ee_Link_Suction_Cup_Suction_Cup",
            maximum_substep_size=0.00334
        )

        # --- 添加以下代码来开启可视化 ---
        # 开启 RMPFlow 的目标坐标轴显示
        # self.rmpflow.set_visualization_enabled(True) 
        # self.rmpflow.visualize_end_effector_position()
        # 开启碰撞球显示（非常有用于排查为什么“不敢放下”物体）
        self.rmpflow.visualize_collision_spheres() 
        # ----------------------------

        self.articulation_rmp = mg.ArticulationMotionPolicy(robot_articulation, self.rmpflow, physics_dt)

        mg.MotionPolicyController.__init__(self, name=name, articulation_motion_policy=self.articulation_rmp)
        self._default_position, self._default_orientation = (
            self._articulation_motion_policy._robot_articulation.get_world_pose()
        )
        self._motion_policy.set_robot_base_pose(
            robot_position=self._default_position, robot_orientation=self._default_orientation
        )
        return
    
    def reset(self):
        mg.MotionPolicyController.reset(self)
        self._motion_policy.set_robot_base_pose(
            robot_position=self._default_position, robot_orientation=self._default_orientation
        )