You must give it sufficient consideration and properly organize the signals to easily pass the testing phase.

Guide to the Gateway to the Next Level of ESP32-Based Advancement Through Concrete Practical Training in the Curriculum

[Editor's Note] Edge AI is a technology that reduces latency and security issues by processing data directly on the device, and the ESP32 enables the implementation of practical AI through the selection of optimized models and boards, even within limited memory and computational performance. For the development environment, one selects between the simplicity of Arduino and the expertise of ESP-IDF according to the purpose, while the core of the KiCAD practice is to acquire a sense of circuits and layout through iterative fabrication. Accordingly, on May 22, we met with Kim Bong-jo, CEO of Care Lab, who is conducting the 'Offline Practical Training to Design and Implement ESP32-Based Edge AI IoT Products' as an offline practical session at the e4ds Electrical and Electronic Lifelong Education Center, and heard about the training.

Please briefly summarize the technical differences between Edge AI and Cloud AI, and explain the major constraints and solutions when applying Edge AI to the ESP32 family.

'Edge AI' refers to combining edge computing and artificial intelligence to perform machine learning tasks directly on interconnected edge devices.

Simply put, Edge AI, also known as 'On-Device AI,' is a distributed computing paradigm where AI algorithms run directly on devices using data generated by the devices.

Running AI at the edge of a local network is a method that eliminates the need to connect devices to a centralized server, such as the internet or the cloud.

■ How hardware characteristics of boards such as ESP32-S3 and ESP32-CAM (memory, camera interface, hardware acceleration, etc.) affect the actual Edge AI pipeline

Essentially, the data to be processed, device performance, and price are mutually limiting factors.stay.

High-priced and high-performance onboard computers and AI processors are cumbersome to use in edge computing, and they also present cost issues.

As an alternative to overcome these mutual constraints, the performance of the ESP32 is optimized.

It is important to examine the ESP32 family and select from the various boards that match the role and performance of the device to be implemented.

In the process of implementing Edge AI, the role of the board is not about the accuracy of distinguishing right from wrong, but about the clarity of demonstrating the ability to properly perform the assigned tasks.

More importance should be placed on whether the function is performed properly rather than on accuracy.

There is definitely a solution.

■ What are the pros and cons of Arduino IDE and ESP IDF from a practical perspective, and how should the two environments be allocated for teaching in education?

In ESP32 development, Arduino is a beginner-friendly environment focused on rapid prototyping and ease of use, while ESP-IDF (Espressif IoT Development Framework) is Espressif's official professional development environment for hardware control, performance optimization, and advanced features.

Arduino is easy but limited, while ESP-IDF is complex but offers a high degree of freedom.

※ Key Comparison Summary align="center" border="1" cellpadding="1" cellspacing="1" style="width:900px;"> characteristic Arduino ESP-IDF(Espressif) Target Beginner, Hobby, Quick Prototype Experts, commercial products, complex projects Development language C++ (Arduino library) C/C++ (FreeRTOS API, HAL) Learning curve Low (Very Easy) High (Difficult) Hardware control Limited (Simplified API) Total control (low level)/register) Build/Configuration Arduino IDE (Simple) CMake/VS Code (complex) Library Very rich (community) Official components, relatively few
- Choose Arduino: For those new to ESP32, when you want to quickly build a prototype within 1-2 days, or for projects where libraries are important.

- ESP-IDF Selection: When developing commercial products, when utilizing deep sleep mode to the extreme, or when complex multitasking using multicore is required.

■ When applying lightweight models such as YOLO or Edge Impulse on an ESP32, what techniques are recommended in terms of model reduction, quantization, inference speed, and memory management?

It does not cover such in-depth content.

For example, in the case of video object analysis using a camera and a microcontroller with 16MB of flash memory and 8MB of RAM, the process involves transmitting camera footage to a PC and utilizing the YOLo algorithm for image object analysis. Additionally, by undergoing the image training process provided by Edge Impulse (Qualcomm), a library optimized for edge devices is generated for the boardThere is a way to upload and run it.

Recently, ESP-Claw, an AI agent framework specifically designed for ESP32 boards, was released, so I am also watching it with interest.

This framework is said to implement a 'Chat Coding' paradigm that defines device behavior through natural language conversation and executes it directly within the chip.

I am curious about the core circuit and layout principles taught in the module design practice using KiCAD. Also, what precautions should be taken during the manufacturing and testing phases?

First of all, I assume that most embedded developers have also worked with onboard computers such as Arduino, ESP32, and Raspberry Pi.

He is also expected to excel in firmware development and device debugging.

The final destination is the EDA tool.

If you can easily design circuits and manufacture PCBs using the open-source KiCAD to create boards, you are invincible.

Initially, test with a Dev Module, and then completely unpack the circuit and design it using an EDA tool.

In fact, it is decided before manufacturing.

Only by giving sufficient consideration and properly organizing the signals can you easily pass the test phase.

There are no special principles or rules. After going through PCB fabrication and assembly two or three times, you will learn the lesson that there is much to learn and that you have a long way to go.

■ Is there any preparation you would recommend first to engineers considering this training?

If you have a rich imagination and enjoy a new journey where results are important, it does not matter whether you are new to hardware coding, already working as a developer, or an educator in maker activities.

In fields of making things, where one uses both head and hands to create and develop, the steps one must take do not become a major obstacle.

It means that some hurdles can simply be ignored.

Having extensive foundational knowledge can be helpful, but it can also lead to making premature judgments that limit many possibilities.

The curse of knowledge works everywhere.

The ESP32 IoT Product Development Hands-on Training allows you to experience the exciting journey of embedded, maker, and digital product development, and guides you to the door to taking a leap forward based on the ESP32 through concrete hands-on practice.

Of course, you are the one who opens the door.