Five core curricula, spanning fromembedded AI trainingtothe OpenArm open-source armthen toembodied AI robots、dexterous handsandOpenDuck Mini— a full-chain talent pipeline for new-engineering programs.
Built on the WWNT training kit hardware, spanning fromlow-level systems, embedded development, and computer vision to edge AI deployment— a closed-loop teaching path. 6 core courses, 100+ lab projects, covering 36 interdisciplinary majors.
Powered by a Raspberry Pi 5 terminal with strong compute for complex algorithms and a stable embedded base.
Plug-and-play temperature/humidity, light, and ultrasonic sensors covering the full IoT sensing layer.
Driver-free camera, touchscreen, and microphone for computer vision and human-robot interaction labs.
Modules snap together magnetically, building scenarios from simple control to complex AIoT with a low setup barrier.
Linux OS · sensor technology
Building the hardware and system base
Embedded system development
System architecture and driver development
Computer vision · image processing
Algorithm validation and engineering practice
Embedded AI applications
Algorithm + hardware closed loop
The cornerstone of embedded development — master Linux architecture, core operations, and advanced administration, from command line to kernel and embedded adaptation.
System architecture and distro differences, terminal, filesystem, Vim, pipes and I/O redirection
Process lifecycle, network config, systemd services, RPM/DEB packages, environment variables
Cross-compilation, ARM toolchain, kernel boot, root filesystem, shell scripts and autostart
Covers common sensor principles and applications, building multi-sensor acquisition, processing, and fusion skills for a full perception-processing-transmission chain.
Photoelectric/temp/humidity/inertial sensors, AD conversion and digital filtering, I2C/SPI/UART protocols
Environmental sensing, attitude calculation, camera/IR/ultrasonic experiments
Linux sensor drivers, multi-device sync acquisition, MQTT/HTTP cloud upload
Focuses on core embedded technology, covering the full flow from hardware architecture to software implementation, building low-level driver, peripheral debugging, and complex application skills.
ARM architecture, MCU/MPU differences, GPIO/UART principles and register operations
Linux character device drivers, I2C/SPI frameworks, interrupt and timer drivers
C embedded programming, multithreading and synchronization, network socket programming
Introduces classic computer vision algorithms and OpenCV, combining theory with practice to build image processing and feature analysis skills for real projects.
Image representation, color space conversion, filtering, edge detection, thresholding
Geometric transforms, SIFT/ORB feature matching, object and scene matching
OpenCV engineering, ARM/Linux cross-compilation, real-time video and algorithm optimization
Deep application of vision algorithms in industry and daily life, turning core algorithms into deployable solutions through real projects.
Industrial preprocessing, denoising/correction/illumination, semantic/instance segmentation and subpixel detection
Barcode and OCR, defect detection, keypoint localization and part pose analysis
ROI extraction and NEON acceleration, dynamic frame rate and compute allocation, low-power operation
The capstone of the AIoT curriculum — deep learning from algorithm to hardware, covering training, lightweight optimization, and edge deployment.
CNN, MobileNet/YOLO-Nano lightweight models, quantization and pruning
Industrial dataset annotation, PyTorch→ONNX→Tengine/NCNN conversion
Linux edge inference, real-time camera + AI, GPIO/display closed loop
Deploy lightweight YOLO to Raspberry Pi 5 for real-time detection and GPIO integration
A fully open-source domestic humanoid arm built on open hardware and ROS2, focused on cross-disciplinary practice. Five categories:Foundations, Core Majors, AI Integration, Mechanical Design, Integrated Practice, covering core undergraduate and vocational courses.
Vocational / lower-division
Advanced undergraduate / vocational
Undergraduate / AI track
Mechanical majors
Campus electives · innovation
Based onthe Mini Pi embodied robothardware platform, spanning fromenvironment setup, device debugging, and interaction control to simulation and deployment— a closed-loop practice path with 10 core lessons (54 hours) in basic (1–5) and advanced (6–10) modules, plus competition, resources, faculty, and industry-research support.
Set up the dev environment — simulation platform, dependency config, and toolchain — laying the software and system foundation.
Robot hardware deployment and debugging — sensor calibration, actuator calibration, and comms testing, from unboxing to running.
A flagship voice-interaction lesson covering ASR/TTS APIs and multi-turn dialogue to build full voice interaction.
Hands-on tool integration — RESTful APIs, sensor data reading, and command dispatch for robot-system interconnection.
Basic motion command programming — joint control, motion sequencing, and kinematics, from single to multi-joint coordination.
Build simulation scenes in Isaac Sim / NVIDIA Omniverse with asset import, providing a digital base for RL training and algorithm validation.
Hands-on drag-teaching — motion recording, trajectory replay, and teach programming, enabling zero-code robot motion learning.
A flagship RL training lesson — policy training, domain randomization, and Sim2Real transfer, from virtual to real.
Deploy simulated models to real robots — conversion, firmware flashing, and testing — closing the sim-to-real loop.
An integrated project — perception → planning → execution — using robot soccer to tie together all prior lessons and validate the full tech stack.
Focused on computer application and digital literacy competitions, with WorldSkills embodied intelligence coaching through real cases and teamwork.
A full resource library — syllabus, lesson design, per-chapter videos, and source code, with workbook and project-based textbooks for blended teaching.
Holiday training on cutting-edge tech, deep teaching and curriculum design, building dual-qualified faculty through hands-on practice.
Guide teaching reform and industry-education projects, landing MOE partnerships with real enterprise projects and data annotation.
WWNT Robotics · based on LinkerHand ROS SDK a tiered curriculum for vocational, undergraduate, and graduate levels. Built on the full LinkerHand ROS1/ROS2 SDK and wwntAI integration, delivering a three-tier standardized dexterous-hand teaching system.
SDK reference:LinkerHand ROS SDK docs | View dexterous hand details
48 hours · industrial robotics / mechatronics
Objectives:Master dexterous hand assembly, GUI control, Python basics, and industrial grasping to debug devices and develop simple automation stations.
Hardware:L6/O7 low-cost hands, desktop bench, wwntAI visual control software
64 hours · robotics / automation / AI
Objectives:Master ROS1/ROS2, multimodal fusion, hand-eye coordination, simulation, and teleoperation to independently integrate systems and design courses.
Hardware:L10/L16 hands, PyBullet workstation, monocular camera, wwntAI cluster system
32 hours · elective / research
Objectives:Use the LinkerHand SDK for imitation learning, grasp generation, and embodied model deployment, supporting theses, competitions, and grants.
Hardware:L20 flagship 21-DOF hand, humanoid robot, GPU workstation, DexScale tactile dataset license
For new electives — L6/L10 hands + full SDK source + tiered lab guides + wwntAI software + 1 faculty training
For new-engineering programs — 32-hand cluster + sim server + three-tier curriculum + lab construction + 2 years on-site service
For graduate/competition — L20 flagship + deep learning workstation + full imitation/grasping source + joint research + full competition support
| SDK case ID | Teaching module | Core knowledge point |
|---|---|---|
| 0000-PyBullet simulation | Undergraduate sim module | Offline simulation, sim-real code interop |
| 0001 - Get hand state | Undergraduate ROS dev | ROS topics, joint data parsing |
| 0002-gui_control GUI | Vocational basics | Visual control, waypoint recording/replay |
| 0003~0005 force/speed/current | Vocational grasping, undergraduate fusion | Multi-sensor acquisition, adaptive grip force |
| 0101-lipcontroller tactile pinch | Vocational integrated | Tactile-triggered closed-loop grasping |
| 0102~0106 Python gestures | Vocational Python | Python ROS nodes, continuous motion control |
| 0201~0203 L25 teleoperation | Undergraduate teleoperation | Remote control, multi-mode switching |
| 0009-finger_guessing rock-paper-scissors | Undergraduate hand-eye | Vision recognition + hand interaction loop |
| 1001-human-dex imitation learning | Graduate research | Teaching data acquisition, RL training |
| 1002-linker_unidexgrasp grasping | Graduate research | General grasping network, cross-model pose mapping |
WWNT Robotics · OpenDuck Mini companion courses, coveringembodied intelligence → bipedal motion control → MuJoCo simulation— a full chain using the ~42 cm, 14+2-DOF open bipedal robot, from theory to real robot, simulation to RL, with zero barrier.
Related product:View OpenDuck Mini details | Open-source repository
Course objectives, knowledge map, learning path, and assessment overview.
Configure the dev environment, SSH, and file sync toolchain.
From traditional AI to embodied intelligence: the perception-decision-action loop.
Layered control: high-level planning, mid-level motion, low-level servo.
Model-based vs. learning-based control — comparison, use cases, and evolution.
COM, support polygon, and static stability geometry.
Linear inverted pendulum model and capture point computation.
Single/double support transitions, COM trajectory planning, stability criteria.
DH modeling, forward kinematics, numerical inverse kinematics.
Foot trajectory planning: lift-off, swing, and touch-down constraints.
Predicting swing-foot touch-down using COM velocity and position.
From state machine to gait cycle: decomposing bipedal walking.
Parameter tuning, online debugging, and real-robot gait validation.
MuJoCo physics, URDF/MJCF models, and contact models.
Install MuJoCo on Linux/Raspberry Pi and bind Python.
Load the OpenDuck Mini model, run the first step, and visualize.
A ~42 cm bipedal robot with 14+2 DOF and 35+ fully open 3D-printed parts, ready out of the box.
MuJoCo simulation + pre-trained RL walking policies, full Sim2Real for research and teaching.
Custom colors, badges, and mascots to create a campus-exclusive AI ambassador.
Cultivating well-rounded embodied-intelligence talent with solid theory, strong engineering practice, and cutting-edge innovation.
Full-stack development from hardware selection and drivers to software architecture and integration
Master deep learning and RL, applying them to perception, decision, and control tasks
Turn theory into practice — debug hardware, optimize code, and solve real-world problems
Agile collaboration and cross-functional communication to achieve R&D goals
The WWNT curriculum provides complete solutions from hardware to content for new-engineering programs. Contact us for the full syllabus and implementation plan.