Explore the ZEKEEP ZK-EM600 6-axis embodied AI robotics learning platform for robotics education and research, featuring ROS 2, MoveIt 2, master-slave teleoperation, RGB-D vision, motion planning and AI-assisted grasping.
Payload
0.5 kg
Reach
750mm
Repeatability
±1mm
End Speed
-
From Mechanical Assembly to AI-Assisted Grasping — A Structured Learning Path for Robotics Development and Embodied AI.
The ZEKEEP ZK-EM600 6-Axis Embodied AI Robotics Learning Platform is an integrated learning platform designed for robotics education, intelligent manufacturing training and practical exploration of embodied AI technologies.
Built around the ZK-EM600 six-axis robotic arm, the platform combines master-slave teleoperation, robot control, ROS 2, motion planning, 3D vision and AI-based grasping technologies.
It helps learners progress from mechanical structure familiarization and assembly calibration to robotics software development, visual perception and intelligent task execution.
Based on physical robotic hardware, the platform works with a software environment including Ubuntu 22.04, ROS 2 Humble and MoveIt 2. Practical topics include robot modeling, joint control, kinematics, trajectory planning, collision detection and real robotic arm execution.
Learners can start with fundamental robot control and gradually understand the complete technical workflow from simulated motion planning to real hardware operation.
For perception and grasping applications, the platform documentation presents an RGB-D 3D vision workflow based on the Orbbec Gemini 2. Topics include color images, depth maps, point clouds, 3D coordinates and Eye-in-Hand hand-eye calibration.
These capabilities can be combined with YOLOE object recognition, instance segmentation and GraspNet 6-DOF grasp pose estimation to explore the processes involved in target perception, spatial positioning and grasp pose generation.
The platform also includes master-slave teleoperation components for experiments involving leader-arm operation, joint mapping, follower-arm tracking and drag teaching.
For further development and research, learners can explore imitation learning, VLM / LLM Agents, ACT / VLA and other embodied AI technologies.
Unlike robotic teaching equipment designed primarily for basic motion control, the ZK-EM600 learning platform emphasizes the relationship between mechanical structures, software control, visual perception, motion planning and intelligent execution.
It is suitable for vocational education, university robotics laboratories, innovation projects and related research and teaching environments.
Through a structured learning path, students can develop an integrated understanding of robotic arm assembly, robot programming, vision-guided grasping and embodied AI applications, establishing a foundation for further robotics development and intelligent automation research.
Understand robot motion principles through real mechanical hardware and joint control.
The platform uses the ZK-EM600 six-axis robotic arm, with a reference rated payload of 0.5 kg and a maximum reach of 750 mm. The robotic arm itself weighs approximately 5.2 kg.
Learning activities cover six-axis structure familiarization, mechanical assembly, zero-position calibration, joint communication and basic motion control.
These activities help students understand the relationship between mechanical structures and robot motion control, providing a foundation for further robotics development and integrated applications.
Connect robot modeling, motion planning and physical robotic arm execution.
The software learning pathway includes ROS 2 Humble, URDF / Xacro, ros2_control, RViz 2 and MoveIt 2.
Students can explore robot modeling, node communication, joint control, kinematics, path planning, collision detection and trajectory execution.
These practical activities help explain how robotics software interacts with physical hardware.
Plan and validate movements before applying them to the physical robot.
The documented learning workflow includes robot motion planning, environment modeling, collision detection, virtual motion preview and real robotic arm execution.
By comparing simulated results with actual operation, students can understand the relationship between planned paths, robot states and physical motion, developing practical skills in robotic system debugging and problem analysis.
The leader arm provides motion input while the follower arm executes the corresponding movements.
The platform includes a six-degree-of-freedom master-slave teleoperation leader arm.
Documented functions include zero-position alignment, joint mapping, joint limit mapping, gripper synchronization and drag control.
Subject to the corresponding hardware and software configuration, learners can conduct robotic motion teaching, master-slave tracking, trajectory recording and playback experiments.
These functions provide a foundation for studying robot teleoperation and demonstration data collection.
Move beyond 2D images to depth information and 3D point clouds.
The platform's vision learning modules focus on RGB-D data processing, including RGB images, depth maps, point clouds, XYZ coordinates and image-depth alignment.
Using tools such as OpenCV and Open3D, students can explore image processing, point cloud generation, spatial coordinate extraction and vision data analysis.
These activities help learners understand the fundamentals of robotic visual perception.
Transform visual recognition results into spatial poses that robots can use.
The platform documentation presents coordinate relationships involving the camera, robot base and gripper, together with an Eye-in-Hand calibration workflow.
Through multi-pose data collection, coordinate transformation calculations and RMSE-based accuracy evaluation, students can learn how to establish spatial mapping between a camera and robotic arm.
This provides a foundation for applications such as vision-guided grasping.
From object recognition to 6-DOF grasp pose estimation.
The documented technical workflow includes YOLOE object recognition, instance segmentation, point cloud processing and GraspNet 6-DOF grasp pose estimation.
Learners can explore object detection, target region extraction, spatial information processing and candidate grasp pose evaluation.
These activities help explain the process of generating robotic grasping actions from visual perception.
Actual functions, recognition performance and algorithm deployment depend on the software version and operating environment.
Build an integrated understanding of robotics, from basic hardware to intelligent applications.
The platform provides a progressive learning pathway covering mechanical assembly, robot modeling, simulation-based planning, physical control, visual perception, hand-eye calibration, AI-assisted grasping and master-slave teleoperation.
Different technical modules support different learning tasks, helping students connect mechanical structures, robotics software, vision algorithms and motion execution.
Explore task understanding and intelligent robotic execution beyond basic motion control.
Building on fundamental robot control and vision-based grasping, the platform documentation also introduces embodied AI technologies such as VLM / LLM Agents, LeRobot, ACT and VLA.
Subject to the actual development environment, students can explore task understanding, action planning, demonstration data collection and imitation learning.
These technologies provide additional learning and research directions for embodied AI and intelligent robotics applications.
It is designed for practical learning and research involving robotic structure familiarization, mechanical assembly, robot control, ROS 2 development, motion planning, 3D vision, hand-eye calibration, master-slave teleoperation and AI-assisted grasping.
The platform uses the ZK-EM600 six-axis robotic arm.
Its reference specifications include a rated payload of 0.5 kg and a maximum reach of 750 mm.
The documented software environment includes Ubuntu 22.04, ROS 2 Humble, MoveIt 2, URDF / Xacro, ros2_control and RViz 2.
These technologies support practical learning in robot modeling, motion planning, control and visualization.
Actual installation status and delivered software versions should be confirmed against the official configuration.
Real robotic arm control is one of the main learning directions of the platform.
Students can explore the workflow from motion planning and trajectory validation to physical robot execution.
Actual executable functions depend on the controller software, hardware interfaces and delivered configuration.
The documentation includes RGB-D depth data, point cloud processing, XYZ coordinate acquisition and Eye-in-Hand hand-eye calibration.
These technologies can be used for practical vision experiments and development.
Specific software functions and ready-to-run demonstration programs should be confirmed against the delivered software version.
The product documentation presents technical workflows involving YOLOE object recognition, instance segmentation, point cloud processing and GraspNet 6-DOF grasp pose estimation.
Actual recognition, pose estimation and grasp execution capabilities depend on software deployment, hardware configuration and target object conditions.
Master-slave teleoperation captures operator movements through a leader arm and maps the corresponding joint motions to a follower arm for execution.
This technology can be used for robot motion teaching, master-slave control and related practical research.
It is suitable for robotics engineering, industrial robotics technology, mechatronics, intelligent manufacturing, automation and artificial intelligence programs.
Actual course suitability should be determined according to the institution's curriculum, training projects and available software development access.
The documentation presents related technologies including VLM / LLM Agents, LeRobot, ACT and VLA as potential learning and research directions.
Actual model deployment, training, inference and data collection capabilities depend on the delivered software and hardware environment.
To better match the platform with specific teaching and research requirements, we recommend clarifying the following points before selection:
Course Focus: Robotic arm control, ROS 2 development, motion planning, 3D vision, AI-assisted grasping or embodied AI research.
Hardware Configuration: Specific options for the six-axis robotic arm, gripper, master-slave teleoperation leader arm, vision equipment and power supply.
Software Requirements: Installation status, versions and development access for ROS 2, MoveIt 2, vision processing and AI algorithms.
Project Requirements: Whether physical robotic arm execution, master-slave teleoperation, hand-eye calibration, point cloud processing or AI grasping functions are required.
Development Extensions: Whether additional work is planned involving demonstration data collection, imitation learning, VLM / LLM Agents or ACT / VLA research.
By clarifying course objectives, experimental projects and equipment requirements, the hardware configuration, software environment and training content can be better matched to the intended application.
| Parameter | Specification |
|---|---|
| Product Name | 6-Axis Embodied AI Robotics Learning Platform |
| Robotic Arm Model | ZK-EM600 |
| Degrees of Freedom (DOF) | 6 DOF + Gripper |
| Rated Payload | 0.5 kg |
| Maximum Reach | 750 mm |
| Repeatability | Original image indicates +1 mm; confirmation required as to whether the specification should be ±1 mm |
| Robotic Arm Weight | 5.2 kg |
| Installation Options | Floor-mounted / Desktop-mounted |
| Power Supply | 24V Regulated Industrial Power Supply |
| Joint Configuration | Six Joints, J1 through J6 |
| Parameter | Specification |
|---|---|
| End Effector | Two-Finger Parallel Electric Gripper |
| Communication Method | Serial-to-CAN Adapter Module |
| Motion Control Capabilities | Robot joint control, motion planning, trajectory execution and grasping operations |
| Application Areas | Robot structure familiarization, assembly, control, motion planning, vision applications and embodied AI development |
| Joint | Motion Range |
|---|---|
| J1 (Base) | -148° to +148° |
| J2 (Shoulder) | 0° to +212° |
| J3 (Elbow) | 0° to +212° |
| J4 (Wrist) | -90° to +90° |
| J5 (Wrist) | -90° to +90° |
| J6 (End Joint) | -90° to +90° |
The actual joint motion ranges are subject to the technical specifications of the delivered robotic arm.
The leader arm is used to capture operator movements and provide motion input to the follower arm. It is an important component for master-slave control, robotic teleoperation and motion teaching.
| Parameter | Specification |
|---|---|
| Device Name | Master-Slave Teleoperation Leader Arm |
| Control Role | Leader (Master Arm / Operator Input Device) |
| Number of Joints | 6 DOF |
| Power Supply | 12V |
| Zero-Position Alignment | Supported |
| Joint Mapping | Supported |
| Communication | USB Serial Communication |
| Joint Limit Mapping | Supported |
| Gripper Synchronization | Supported |
| Drag Teaching | Supported |
| Main Applications | Master-arm motion input, joint mapping, follower-arm tracking and teleoperation experiments |
Basic Master-Slave Teleoperation Workflow:
The operator provides motion input through the leader arm. The control system maps the leader arm's joint movements to the follower arm, which then executes the corresponding movements.
Actual mapping relationships, control frequency, tracking accuracy and communication latency depend on the control system and software configuration.
| Parameter | Specification |
|---|---|
| Operating System | Ubuntu 22.04 |
| Robotics Middleware | ROS 2 Humble |
| Motion Planning Framework | MoveIt 2 |
| Robot Description | URDF / Xacro |
| Robot Control Framework | ros2_control |
| Kinematics and Trajectory Planning | KDL / TRAC-IK, MoveIt 2 / OMPL |
| Programming Languages | Python / C++ |
| Robot Model Visualization | RViz 2 |
| Web-Based 3D Visualization and Control | Three.js / rosbridge |
| Vision Data Types | RGB Images, Depth Maps and Point Clouds |
| Vision Processing Tools | OpenCV / Open3D |
| Object Recognition | YOLOE |
| Instance Segmentation | Mask-Based Instance Segmentation Workflow |
| 6-DOF Grasp Pose Estimation | GraspNet 6-DoF |
| Motion Safety Functions | Documented workflows include inverse-kinematics reachability checks, collision detection, TCP height checks and virtual motion previews |
Specific software package versions, source code availability, delivered functions and accessible interfaces are subject to the official delivery specification.
| Component | Function |
|---|---|
| Six-Axis Robotic Arm | Provides six degrees of freedom for trajectory execution, object grasping and practical robot tasks |
| Two-Finger Parallel Electric Gripper | Used for workpiece gripping and pick-and-place operations |
| Master-Slave Teleoperation Leader Arm | Captures operator movements and supports joint mapping and teleoperation control |
| J4340Px3 Component | Included in the robotic arm kit |
| J4310x4 Component | Included in the robotic arm kit |
| Serial-to-CAN Adapter Module | Provides serial-to-CAN communication adaptation for compatible devices |
| 24V Regulated Industrial Power Supply | Supplies operating power to the robotic arm hardware |
| Woodworking Clamp | Component listed in the hardware configuration illustration |
| Set of Building Blocks | Used for robotic grasping, material handling, object recognition and sorting exercises |
The above list describes the documented hardware components and does not necessarily indicate that every item is included as standard equipment. Please confirm the standard configuration and optional components with the sales team before ordering.
| Functional Module | Main Functions and Technologies |
|---|---|
| Robotic Arm Assembly and Calibration | Structural familiarization, mechanical assembly, motor communication, zero-position calibration and joint parameter adjustment |
| ROS 2 Robot Control | Node / Topic, Service / Action, ROS 2 control and joint state management |
| Robot Modeling | URDF / Xacro robot description and structural visualization |
| Motion Planning | MoveIt 2, kinematics, path planning and trajectory optimization |
| Simulation and Collision Avoidance | Planning scene configuration, collision detection, dynamic obstacle avoidance and virtual motion preview |
| Real Robot Execution | Motion plan transmission, physical robotic arm control and task execution |
| Master-Slave Teleoperation | Leader-arm input, joint mapping, follower-arm tracking and drag teaching |
| RGB-D 3D Vision | Color images, depth maps, point cloud generation and XYZ coordinate acquisition |
| Hand-Eye Calibration | Eye-in-Hand calibration, camera-to-gripper coordinate transformation and calibration accuracy evaluation |
| AI-Based Grasping | Object recognition, instance segmentation, point cloud processing and 6-DOF grasp pose candidate generation |
| Embodied AI Development | VLM / LLM Agents, LeRobot, ACT / VLA and related learning and research directions |
Maximum Robotic Arm Reach: 750 mm
Robotic Arm Weight: 5.2 kg
Rated Payload: 0.5 kg
Installation Options: Floor-mounted / Desktop-mounted
Robotic Arm Power Supply: 24V Regulated Industrial Power Supply
The above dimensions and weight refer to the robotic arm itself, not the complete learning platform.
Suitable for practical training in industrial robotics, robotics engineering, mechatronics, intelligent manufacturing and automation programs.
Students can begin with mechanical structure familiarization, robotic arm assembly, zero-position calibration and joint control, then progress to ROS 2, robot modeling, trajectory planning, real robot control and integrated task execution.
Suitable for courses and laboratory projects involving robot control, kinematics, robotics middleware, 3D vision, hand-eye calibration and intelligent grasping.
By combining ROS 2, MoveIt 2, RGB-D vision and AI-based grasping technologies, students can explore the relationship between basic robot control, perception, planning and execution through structured teaching and project-based practice.
Suitable for experiments involving embodied AI, visual perception, robotic manipulation and intelligent task execution.
Subject to the available software, hardware and model configurations, students can explore task understanding and planning using VLM / LLM Agents, combining robot control, visual perception and grasping execution to study the workflow from task descriptions to physical robotic actions.
The inclusion of related learning frameworks does not necessarily mean that all large language models or embodied AI algorithms are pre-deployed. Actual available functions depend on the delivered configuration.
Suitable for robotics software development and system integration training.
Practical topics include ROS 2 node communication, topics, services, action interfaces, URDF / Xacro modeling, ros2_control and MoveIt 2 motion planning.
Students can begin with robot modeling and simulation before progressing to physical robotic arm control, developing practical skills in robotics software development and system debugging.
Suitable for RGB-D vision processing, point cloud generation, coordinate transformation, hand-eye calibration and vision-guided grasping.
Using RGB images, depth maps and point cloud data, students can explore the relationship between 2D image information and 3D spatial data, and learn how visual recognition and spatial positioning results can be converted into executable robotic movements.
Suitable for master-slave robot control, motion teaching, joint mapping and robotic manipulation research.
Through motion mapping between the leader and follower arms, students can understand operator motion capture, joint mapping, synchronized control and follower-arm execution.
Under the corresponding system configuration, further experiments can explore teleoperation data collection and imitation learning.
Suitable for interdisciplinary innovation projects combining mechanical structures, motion control, software development, 3D vision and AI algorithms.
Depending on teaching conditions and student experience, projects can cover robotic picking and material handling, motion planning and obstacle avoidance, visual positioning, master-slave teleoperation and AI-assisted task execution.
Study the six joints J1 through J6, including the base, upper arm, forearm, wrist and end effector.
Learn mechanical connections, assembly sequences, positioning and fastening, and overall structural inspection.
Understanding the joints and their motion relationships provides a foundation for subsequent studies in robot kinematics, control and path planning.
Learn the fundamentals of motor communication, joint parameter reading, CAN communication adaptation and zero-position calibration.
Through zero-position setting, data storage, communication checks and basic operational verification, students can understand the commissioning and calibration procedures required after robotic arm assembly.
Actual wiring and calibration procedures must follow the equipment technical documentation.
Learn the core concepts of ROS 2 Nodes, Topics, Services and Actions.
Through tasks such as reading joint states, sending control commands and receiving motion feedback, students can understand data exchange between different robotics software nodes.
Combined with ros2_control and the available control interfaces, students can progressively perform robotic arm joint control experiments.
Use URDF / Xacro to describe robotic links, joints, motion relationships and overall robot structure.
Further visualize the robot model using RViz 2.
Learn robot structure description, coordinate frame definition and joint motion relationships, preparing for kinematics calculations and subsequent simulation-based planning.
Use MoveIt 2 and planning components such as OMPL to learn how to configure robot starting states, target states, planning scenes and motion paths.
By examining planned trajectories, adjusting target positions and comparing different planning results, students can understand how robots plan movements from an initial state to a target state.
Add obstacles to simulation or planning scenes to conduct collision detection, path adjustment and obstacle-avoidance experiments.
By comparing the feasibility and execution results of different paths, students can understand environmental constraints, collision checking and motion safety in robot planning.
Actual algorithm performance depends on the planning algorithms and scene configuration.
Integrate motion planning results with physical robotic arm control.
Learn joint commands, trajectory transmission, state feedback and motion execution workflows.
By comparing simulation results with actual robotic movements, students can understand the relationship between robot models, motion planning and physical hardware.
Use the master-slave teleoperation leader arm to provide motion input.
Learn about joint mapping, zero-position alignment, joint limit mapping and synchronized gripper control.
Subject to the available control system, experiments can include leader-arm motion capture, follower-arm tracking, drag teaching, and motion trajectory recording and playback.
Based on the Orbbec Gemini 2 vision device and related software environment, learn the fundamentals of RGB color images, depth maps, 3D point clouds and XYZ spatial coordinates.
Through image-depth alignment, point cloud generation and spatial coordinate acquisition, students can understand how 2D image data is transformed into spatial information for robotic applications.
Study the relationships between the camera coordinate system, image coordinate system, gripper coordinate system and robot base coordinate system.
By collecting calibration data at different robot poses, students can establish the spatial transformation between the camera and gripper, and evaluate calibration error using metrics such as RMSE.
This experiment helps students understand how visual recognition results can be transformed into target poses for robotic execution.
Explore object recognition and instance segmentation workflows using YOLOE.
Learn object detection, category identification, object region extraction and recognition result processing.
Combined with actual materials and visual environments, students can explore how recognition results are applied to robotic grasping tasks.
Supported object categories, open-vocabulary capabilities and runtime performance depend on the software version and deployment environment.
Based on point cloud processing and GraspNet 6-DOF grasp pose estimation workflows, learn how to generate and evaluate candidate grasp poses.
By considering target position, gripper orientation and robot kinematic constraints, students can understand how visual perception generates pose information for robotic grasping.
Before executing a physical robot task, conduct virtual motion previews of grasp poses and planned paths.
Depending on the supported system functions, verify inverse kinematics reachability, collision risks and TCP height constraints.
By comparing virtual planning with actual execution, students can understand safety checks and motion validation in robotic grasping tasks.
Explore robot state visualization, 3D model rendering and web-based interactive control using technologies such as Three.js and rosbridge.
Subject to the actual software access and available interfaces, projects can cover browser-based robot status monitoring, motion control and experiment visualization.
Building on basic robot control, visual perception and grasping workflows, students can further explore embodied AI technologies such as VLM / LLM Agents, LeRobot, ACT and VLA.
Research topics may include task understanding, action planning, demonstration data collection, imitation learning and robotic task execution.
Model training, inference deployment and data collection capabilities must be evaluated according to the actual hardware and software environment. The inclusion of these technologies as learning directions does not mean that all functions are pre-installed or fully implemented.
Leave your contact information and requirements. Based on the product specifications and application scenarios, we will provide preliminary selection and solution advice.
