The ZEKEEP Roboarm_v1 is an AI-enabled 3-DOF desktop robotic arm designed for robotics education, AI development, and hands-on programming. It integrates STM32-based motion control with machine vision and voice interaction technologies, providing a practical platform for learning object recognition, task processing, and robotic pick-and-place operations.
Payload
0.5 kg
Reach
~340mm
Repeatability
±1mm
End Speed
-
Vision Recognition × Voice Interaction × Robotic Execution — Enabling a Desktop Robotic Arm to Interpret Tasks and Perform Pick-and-Place Operations.
The ZEKEEP AI-Enabled Interactive Desktop Robotic Arm is a desktop robotics platform designed for artificial intelligence education, robotics programming, and integrated development projects.
Built around an STM32-controlled 3-DOF robotic arm, the platform features a base rotation, shoulder pitch, and elbow pitch configuration. It is equipped with a vacuum suction cup end effector and integrates an industrial PC, USB camera, and supporting software environment to combine robotic arm control with AI applications.
In terms of motion control, the platform follows the basic control architecture of the same-series STM32 desktop robotic arms. It supports practical learning and development involving joint movement, serial communication, target positioning, and robotic action execution.
Reference specifications for the same-series robotic arm include a 0.5 kg rated payload, approximately 1 mm repeatability, and adjustable motion speed. Actual mechanical specifications for the 3-DOF AI version are subject to its specific product configuration.
For visual recognition, the platform combines YOLO series / YOLO-World object detection with OpenCV image processing. It can be used for recognition experiments involving colored blocks, geometric shapes, and other example objects. Through camera calibration and coordinate mapping, learners can explore how image-based target positions are converted into grasping positions within the robotic arm workspace.
For voice interaction, the platform uses FunASR-related speech recognition technology to convert voice commands into text for subsequent task parsing.
For example, when a user gives an instruction such as "Place the red block on the triangle," the system can combine speech recognition results, object detection results, and position mapping workflows to organize the corresponding pick-and-place task.
For software development, the platform documentation lists technologies including Ubuntu 22.04, Python 3.10, ROS 2, OpenCV, and FunASR. These technologies can be used for basic robot control, vision algorithms, voice interaction, and integrated task development.
Subject to the interfaces and project code provided in the actual configuration, learners can progress from individual function testing to integrated robotic interaction projects.
Unlike desktop robotic arms designed solely to execute fixed motions, this platform focuses on the coordination of visual input, voice instructions, and robotic actions. It connects mechanical control, computer vision, speech recognition, and programming, providing an integrated learning platform for AI robotics education and hands-on innovation projects.
Real robotic hardware combined with embedded control.
The platform features a 3-DOF desktop robotic arm consisting of base rotation, shoulder pitch, and elbow pitch joints, driven by stepper motors and an STM32 controller.
Learners can explore joint motion, robotic arm control, serial communication, and basic positioning to understand the relationship between physical robot hardware and control software.
From image acquisition to object detection, providing visual input for grasping tasks.
The platform combines YOLO series / YOLO-World and OpenCV technologies for recognition experiments involving colored blocks, geometric shapes, and other example objects.
By obtaining object categories, pixel coordinates, and related image information, learners can further explore position mapping and robotic grasping control.
Use natural-language instructions to organize robotic tasks.
Using FunASR-related speech recognition technology, the platform converts voice input into text and applies program logic to identify target objects, action types, and placement positions.
For example, users can issue instructions to place a designated colored block at a specified shape position. The task can then be combined with visual recognition and robotic arm control to perform the corresponding operation.
The actual range of supported voice commands depends on the deployed software version and task logic.
Connect visual recognition with physical robotic execution.
The platform documentation includes camera intrinsic calibration and Homography-based coordinate mapping technologies.
Through visual calibration, target positioning, and coordinate transformation, learners can understand the relationship between image coordinates and the robotic arm workspace, establishing a foundation for vision-guided grasping.
Progress from basic positioning to trajectory control.
Documentation for the same-series STM32 desktop robotic arms lists functions such as point-to-point motion, linear motion, linear offset, and palletizing trajectories, as well as joint-coordinate and world-coordinate control.
Subject to the control interfaces retained in the AI version, users can explore basic robotic motion, grasping path adjustment, and target position configuration.
Bring robotic control and software development into one learning environment.
The platform documentation lists software technologies including Ubuntu 22.04, Python 3.10, ROS 2, OpenCV, and FunASR.
By combining basic robot control, visual recognition, and voice-based task processing, learners can understand how different software components exchange data and coordinate tasks.
Build practical skills through progressively structured learning tasks.
The platform supports learning projects covering visual recognition, voice input, task parsing, coordinate transformation, and robotic arm control.
From basic hardware operation to integrated task execution, students can develop a systematic understanding of robotic perception, task processing, and motion control.
This model uses a 3-DOF configuration consisting of base rotation, shoulder pitch, and elbow pitch.
Unlike a four-axis robotic arm, the three-axis version does not include an additional end-effector rotation joint.
The supplied reference specifications list a rated payload of 0.5 kg and repeatability of ±1 mm.
Please refer to the specific mechanical specifications of the AI version for final confirmation.
The same-series product documentation specifies a maximum reach of approximately 340 mm.
The platform is designed around technologies including YOLO series / YOLO-World object detection and OpenCV image processing.
It can be used for recognition experiments involving colored blocks, geometric shapes, and other example objects, providing visual information for subsequent coordinate transformation and grasping tasks.
The platform uses a speech recognition and task parsing workflow based on FunASR.
It supports practical experiments involving voice input, object recognition, grasping, and placement. Actual language support and command capabilities depend on the deployed software version.
Available control functions include point-to-point motion, linear motion, linear offset, palletizing trajectories, and joint-coordinate and world-coordinate control.
Actual availability depends on the current controller program and supported interfaces.
The robotic arm uses a DC 24V / 8A power supply.
The industrial PC, display, and other components of the AI platform may require separate power supplies.
It is suitable for courses involving machine vision, artificial intelligence, robotics programming, automation control, embedded development, and STEM innovation projects.
Its suitability for specific courses should be determined based on the school's experimental projects and teaching objectives.
Available interfaces, project code, software packages, and control permissions are subject to the actual delivery configuration.
To better match the platform to specific teaching and development requirements, the following points should be considered:
Mechanical Performance: Rated payload, maximum reach, horizontal reach, repeatability, and maximum motion speed.
Control System: STM32 controller, USB-to-TTL serial communication, and whether the DI/DO interfaces are included in the AI configuration.
Vision and Voice: Camera specifications, object detection models, FunASR deployment environment, and supported voice instruction range.
Software Development Access: The available modification and development capabilities for Python, ROS 2, vision algorithms, and robotic arm control interfaces.
Equipment Configuration: Whether the industrial PC, display, microphone, camera, calibration board, and experimental materials are included in the standard package.
Deployment Requirements: Actual platform dimensions, weight, power requirements, and required operating space.
By clarifying teaching objectives, experimental projects, and hardware/software requirements, users can determine the appropriate application scope and delivery configuration more accurately.
| Parameter | Specification |
|---|---|
| Product Name | AI-Enabled Interactive Desktop Robotic Arm |
| Model | Roboarm_v1 |
| Robot Structure | 3-DOF desktop robotic arm |
| Degrees of Freedom (DOF) | 3 |
| Joint Configuration | Base rotation, shoulder pitch, and elbow pitch |
| Reference Rated Payload | 0.5 kg |
| Maximum Reach | Approx. 340 mm |
| Horizontal Reach | 340 mm |
| Repeatability | ±1 mm |
| Maximum Motion Speed | Adjustable, up to 120°/s |
| A1 Axis Range | -160° to +160°, with an approximately 40° reserved dead zone |
| A2 Axis Range | -100° to 0° |
| A3 Axis Range | -115° to 0° |
| Drive Motor | Size 42 stepper motor |
| Motor Type | Two-phase open-loop stepper motor |
| Step Angle | 1.8° |
| Rated Current | 2 A |
| Holding Torque | 0.6 N/m |
| Reduction Mechanism | Precision spur-type planetary gearbox |
| Gearbox Backlash | Less than 15 arcmin |
| Controller | STM32 |
| End Effector | Vacuum suction cup; optional gripper depending on the application |
| Reference Digital I/O | 4 DI / 4 DO |
| Robotic Arm Power Supply | DC 24V |
| Reference Power Supply Rating | 24V / 8A |
| Robotic Arm Weight | Approx. 6 kg |
| Base Dimensions | Approx. 220 × 220 mm |
| Operating Temperature | -10°C to +60°C |
| Body Material | 6061 aluminum alloy with anodized finish |
| Base Structure | Sheet-metal base with counterweight structure |
| Parameter | Specification |
|---|---|
| Main Controller | STM32 |
| Basic Control Methods | Joint control, teach-mode control, and serial communication |
| Joint Coordinate System | Basic joint-coordinate control |
| World Coordinate System | World-coordinate control and three-axis coordinated motion |
| Inverse Kinematics | Supported |
| Speed Adjustment | Adjustable speed, up to 120°/s |
| Teaching Point Memory | Supported |
| Maximum Teaching Points | Up to 50 |
| Motion Recipes | Supports 3 sets of operating recipes |
| Safety and Alarms | Safety protection and alarm log functions |
| AI Task Control | Integrates visual recognition, natural-language instructions, and robotic arm control workflows |
| Parameter | Specification |
|---|---|
| Vision Device | USB camera |
| Mounting Components | Camera bracket |
| Connection | USB interface with extension cable |
| Object Detection Technology | YOLO series / YOLO-World |
| Image Processing | OpenCV |
| Object Detection Information | Object categories, bounding boxes, confidence scores, pixel coordinates, and related data |
| Shape Recognition | OpenCV-based contour detection |
| Example Shapes | Triangles, circles, and five-pointed stars |
| Example Objects | Colored blocks, airplanes, cars, bicycles, and other shapes or objects |
| Camera Calibration | Camera intrinsic calibration based on Zhang's calibration method |
| Coordinate Transformation | Homography and related coordinate mapping methods |
| Vision Output | Target positions, recognition results, and related coordinate data |
| Main Applications | Object detection, shape recognition, position estimation, and vision-guided grasping |
Actual camera resolution, recognition accuracy, and frame rate depend on the selected camera and software configuration.
| Parameter | Specification |
|---|---|
| Speech Recognition Framework | FunASR |
| Voice Input | Microphone-based audio capture |
| Speech Recognition Workflow | Voice input, speech-to-text conversion, and instruction parsing |
| Task Parsing | Identifies target objects, action types, placement positions, and other task information |
| Example Task | "Place the red block on the triangle." |
| Visual Information Integration | Combines object recognition and position mapping results |
| Motion Execution | Converts task information into robotic arm pick-and-place actions |
| Main Applications | Voice interaction, task decomposition, and robotic action execution |
Supported languages, complex instruction understanding, offline speech recognition, and recognition accuracy must be confirmed based on the actual software deployment.
| Parameter | Specification |
|---|---|
| Host Computer | Industrial PC |
| Operating System | Ubuntu 22.04 |
| Python Environment | Python 3.10 |
| Robotics Software Framework | ROS 2 (specific distribution to be confirmed) |
| Vision Technologies | YOLO-World, OpenCV |
| Speech Recognition Technology | FunASR |
| Robotic Arm Controller | STM32 |
| Robotic Arm Communication | USB-to-TTL serial communication |
| User Interaction | Voice commands, Python scripts, and web-based interfaces |
| Vision Data Processing | Image acquisition, object detection, position estimation, and coordinate mapping |
| Integrated Tasks | Visual recognition, voice instruction parsing, robotic grasping, and object placement |
| Component | Description |
|---|---|
| 3-DOF Robotic Arm | Performs base rotation, shoulder pitch, and elbow pitch movements |
| STM32 Controller | Handles basic robotic arm control and related communication |
| Vacuum Suction Cup | Grips compatible experimental objects |
| USB Camera | Captures images of the experimental workspace for visual input |
| Camera Bracket | Mounts and adjusts the camera position |
| Industrial PC | Runs vision recognition, voice interaction, and host-side control software |
| Display | Displays vision recognition results, software interfaces, and operating information |
| Microphone | Captures voice commands for interaction |
| USB-to-TTL Serial Cable | Enables communication between the host computer and robotic arm controller |
| Power Adapter | Supplies power to the robotic arm and related equipment |
| Colored Blocks and Shape Boards | Used for color recognition, shape recognition, and pick-and-place experiments |
| Black-and-White Checkerboard Calibration Board | Used for camera calibration and related vision experiments |
| Instruction Manual and Project Code | Supports practical learning and project development; actual delivery is subject to the confirmed configuration list |
The standard configuration and included accessories are subject to the actual sales package.
Maximum Reach: Approx. 340 mm
Horizontal Reach: Approx. 340 mm
Robotic Arm Base Dimensions: Approx. 220 × 220 mm
Robotic Arm Weight: Approx. 6 kg
Installation Method: Desktop mounting
Robotic Arm Power Supply: DC 24V / 8A
The dimensions and weight above primarily refer to the basic robotic arm, not the entire AI interactive platform.
Suitable for programs related to artificial intelligence, industrial robotics, mechatronics, automation, and intelligent manufacturing.
Students can explore robotic joint control, basic STM32 control, Python programming, object detection, voice interaction, and vision-guided grasping, developing foundational knowledge from hardware operation to intelligent control.
Suitable for courses in robotics engineering, artificial intelligence, automation, and intelligent manufacturing.
Using technologies such as ROS 2, OpenCV, speech recognition, and robotic arm control, students can conduct integrated practical exercises involving robot motion control, camera calibration, coordinate transformation, and multimodal interaction.
Suitable for laboratories focusing on machine vision, intelligent robotics, voice interaction, and robotic application development.
The platform supports practical projects involving visual input, natural-language instruction parsing, target positioning, and robotic task execution, helping students understand how AI algorithms interact with physical robotic systems.
Suitable for STEM education, robotics programming, and creative technology projects.
Through tasks such as colored block recognition, geometric shape recognition, voice-controlled grasping, and object placement, students can explore programming, computer vision, and automation control in an intuitive and hands-on way.
Suitable for courses involving speech recognition, visual perception, and coordinated robotic motion.
By combining natural-language instructions with visual recognition data, learners can explore how a robot identifies targets, interprets tasks, and controls the robotic arm to perform corresponding actions.
Suitable for practical learning in Python, ROS 2, embedded control, serial communication, and robot software development.
Subject to the interfaces and project code included in the actual configuration, users can explore basic control programming, vision algorithm development, voice instruction parsing, and system integration.
Basic 3-DOF Robotic Arm Control: Learn the movement relationships between base rotation, shoulder pitch, and elbow pitch, and perform basic motion control exercises.
STM32 and Host Computer Serial Communication: Explore USB-to-TTL serial communication, command transmission, and control feedback to understand data exchange between embedded controllers and host computers.
Robot Home Position and Motion Range Experiments: Understand joint motion ranges, motion limits, and basic calibration methods, and learn how to check operating parameters before running the robot.
Camera Intrinsic Calibration: Use a black-and-white checkerboard calibration board to perform camera calibration and understand the relationship between camera imaging parameters and coordinate systems.
YOLO Object Detection: Detect colored blocks, shapes, and other example objects while learning to retrieve object categories, bounding boxes, confidence scores, and pixel coordinates.
OpenCV Geometric Shape Recognition: Use contour detection to recognize triangles, circles, and five-pointed stars, and obtain corresponding object positions and centroid information.
Image-to-Robot Coordinate Mapping: Combine camera calibration with Homography and related mapping methods to learn how to convert image pixel coordinates into target coordinates within the robot workspace.
FunASR Speech Recognition: Perform voice capture, speech-to-text conversion, and instruction parsing to understand how voice input becomes a source of robotic task instructions.
Natural-Language Task Decomposition: Convert instructions such as "pick up the designated colored block and place it at the specified shape" into target objects, actions, and placement positions.
AI Vision-Guided Grasping: Integrate object detection, coordinate mapping, and robotic arm control to perform target recognition, grasping, movement, and placement.
Robotic Arm Trajectory Control: Subject to the available basic control functions, conduct point-to-point motion, linear motion, linear offset, and related trajectory adjustment experiments.
Web-Based Robot Interaction: Use the available web interfaces to explore browser-based control, status display, and interaction with vision recognition results.
Integrated Vision, Language, and Motion Experiments: Combine speech recognition, task parsing, visual perception, and robotic arm control to explore complete workflows from natural-language input to robotic action execution.
Leave your contact information and requirements. Based on the product specifications and application scenarios, we will provide preliminary selection and solution advice.
