 Back

CocoRobo




“Design, Create, Inspire.”


CocoRobo is an EdTech company I co-founded with my friend Dr. Tony Haiyang Xin in 2016. Based in Hong Kong & Shenzhen, we provide a one-stop solution for STEAM education.


#madewithCocoRobo



Countless creative projects were created by our users using our products, including our engineers & curriculum designers. 



Read more over here 


My mission as the CTO is to create easy-to-use products, which embody the concept of P.B.L. (Project-based Learning) methodology, for teachers and students in K-12 system, in the hope of bridging the gap between cutting-edge technologies and classrooms.

I lead the Research & Development team, it’s been 4 years since the beginning of my startup life. Now we have developed a series of education technology products 


CocoMod™


a set of plug-and-play electronic modules that provide many functions to help users with zero coding or electronic experience to build their projects. The product series has evolved two generations by this far;

The 1st Generation 


Specification:
Physical Size:
Modules in 40*40mm:
All except for the Horizontal Adapting Module
Horizontal Adapting Module: 92*40mm

Processor:
Main Controller:
ATMega32u4 (8-bit AVR Microcontrollers)
WiFi Module:
Espressif ESP8266 (L106 32-bit RISC core)

Input/Output:
UART, Software Serial, Digital & Analog I/O, SPI, I²C, I²S

Display (Screen Module):
RGB OLED Screen with 128*128 px of resolution.
Wireless Features:
Bluetooth 2.0 (Bluetooth Module)
IEEE 802.11 b/g/n Wi-F (Wi-Fi Module)
NB-IoT support LTE Cat NB2 (NB-IoT Module)

Programming Language:
CocoBlockly Arduino Mode or Arduino IDE,
Connect to computer using MicroUSB to USB cable via web browser.

Connector:
Double-slot B2B Connector

Power:
5V DC
The 1st Generation is dedicated to serving as part of our STEAM education solution. The main controller is based on an ATMega32U4 chip; we developed over 15+ sensors and modules to make it compatible with our main controller module. Since it’s an Arduino-based board, user can also embed other third-party modules to our hardware, as long as they can support I²C, SPI or UART communication. Read more in our online help website 


The 2nd Generation: “CocoRobo X”


Specification:
Physical Size:
Modules in 45*45mm: All excepted for Camera Module and Gamepad Module
Camera Module: 20*25mm
Gamepad Module: 108*45mm

Processor:
AI Module: 
Kendryte K210 (RISC-V Dual Core 64bit @ 400MHz)
IoT Module: 
Espressif ESP32 (Tensilica Xtensa LX6 microprocessor @ 240 MHz)

Input/Output:
FPIOA (AI Module Only), UART, GPIO, SPI, I²C, I²S, WDT, TIMER, RTC. Most of them are common ports that are also supported by Arduino ans microbit.

Memory:
AI Module:
16M Flash, 8M SRAM (also has an on-board Micro SD Card slot, able to extend storage up to 16 GiB)
IoT Module:
4M Flash, 520 KiB SRAM

System:
AI Module:
MaixPy (A ported version of MicroPython on K210)
IoT Module:
MicroPython for ESP32

Wireless features:
AI Module:
WiFi 2.4 GHz, 802.11 b/g/n, IPEX antenna support.
IoT Module:
WiFi 2.4 GHz, 802.11 b/g/n, dual-mode Bluetooth.

Vision (AI Module only):
OV2640 (2M Pixles),
Exchangable camera lens: regular lens & wide lens

Display:
1.54 inches TFT RGB LCD Screen with 240*240 px of resolution.

Voice Support:
On-board MEMS Microphone, mono-channel output with 48k sampling rate. mic input with 8k sampling rate.

Machine Learning Model support (AI Module Only):
YOLOv3, TinyYOLOv2, MobileNet
KPU: IEEE754-2008 Standard
APU: Audio Proccessor
FFT: Fast-fourier Transfom accelerator

AI Features (AI Module Only):
Basic OpenMV implementation:
Shape tracking, color blob tracking, QR code tracking, barcode tracking, apriltag tracking, etc.

Machine Learning model:
Face detection and recognition, object detection, hand-written digit recognition, speech recognition (MFCC)

Programming Language:
CocoBlockly X or MicroPython,
Connect to computer using MicroUSB to USB cable via web browser.

Connector:
Single-slot B2B Connector

Power:
3.3V DC
The 2nd Generation is dedicated to serving as part of our AI STEM education solution, and the main controllers of this series are separated into two parts: the I.o.T. module and the A.I. Module. The I.o.T module is based on Espressif ESP32 chip, and the A.I. Module is based on Kendryte K210 chip that can provide 1TOPS computing power in low power mode. Our 2nd Generation product is able to bring many robust features to the K-12 classroom in just 45mm×45mm size through our Blockly programming environment—features like OpenMV implementation, Machine Learning Model Inference, Object Classifier, Speech Recognition, etc.



 We also provide product kits in versatile themes: Music & Light, H.C.I., Robotics, I.o.T., A.I.. All of the product kits come with different types of mechanical parts (Laser cutting board) to help the user to build a complete project.


CocoBlockly™



CocoBlockly is an online visual programming environment that enables users to code through a web browser. Based on Google Blockly, we have developed a various version for four programming languages: 

CocoRobo™ Online Model Training Platform


CocoRobo AI Model Training Platform is dedicated to AI + STEAM education. Our online model training platform provides a one-stop model training process for object tracking purpose on our edge computing module, without building a development environment and eliminating the cumbersome configuration process of environment and tools.



Online model training process:
1. Users use our AI module or camera device to capture the dataset to be labeled
2. Users upload the captured image datasets to our platform and tag them online
3. After the labeling process is completed, upload the dataset and click Start Training
4. View the training process of the model in real-time (including the number of iterations and average loss values)
5. Validate the model at any time, and if you are satisfied, export the model (in *.kmodel format)
6. Put the exported model into the SD card in the AI module, and you can call it through the program

 The backend I designed for the beta version of this platform. I was using Darknet and TinyYOLOv3 for the training part of object tracking model, and nncase for the model conversion.



CocoCloud™



CocoCloud is a cloud platform for data gathering and visualizations.



CocoEducation™



CocoRobo Education is an online education platform for K-12 schools that are using our products and services.



Milestone


We have established collaboration with over 200 schools in the Greater Bay Area across Hong Kong, Shenzhen, Macao, and Guangzhou. Apart from K12 schools, CocoRobo is currently collaborating with NIES (National Institute of Education Sciences) and professors at SCNU to write AI curriculum for elementary and secondary school with products and service from CocoRobo.

We hosted various school-wide robot competitions in Hong Kong with business partners like China Mobile HK. For each competition, approximately 30~40 schools have attended.

At the early stage, we were supported by a government funding called TSSSU from the Chinese University of Hong Kong and Hong Kong S.A.R. Afterwards, we have successfully secured two rounds of Angel funds invested by InnoHK, Saltagen Canada, and Particle X.


  Design, Create, Inspire
If you have any detailed question about CocoRobo, please drop an email to support-hk@cocorobo.cc.