NVDIA Jetson Nano allows high school students to build AI self-driving cars! -AI4kids

AI self-driving car racing competition attracts high school students from north, central and south China to gather for the challenge

AI4kids designed a four-day AI self-driving car practical competition at the Central Science Park Robotics Manufacturing Base during the summer vacation, using NVIDIA Jetson nano edge computing. It attracted young people from Taipei, New Taipei, Hsinchu, Taichung, Yunlin, Kaohsiung, and Penghu to sign up! Led by AI4kids’ professional lecturers and teaching assistants, the students went from building their initial machine learning and deep learning knowledge to becoming familiar with JetBot car hardware assembly, motion control (basic programming, remote control), and intelligent control (obstacle avoidance, object following), and finally being able to independently train the AI ​​visual recognition model and complete the Jetbot self-driving car racing competition!

How to give a racing car AI vision?

The self-driving car competition venue is a square track. The challenge for students is that they cannot use any remote control or program settings, but must let the car complete a lap of the track in the shortest number of seconds through computer vision to win. During the process, a group of two must first complete the track image collection, AI model training and correction, and conduct multiple test verifications and parameter fine-tuning. Most of the trainees were trying to train their own AI models for the first time, so they encountered challenges such as reflections from the track when shooting images, uneven road surfaces, incorrect parameter settings leading to poor vehicle control, etc. They had to make multiple adjustments before the car could run a lap smoothly. After spending half a day training, the trainees all exclaimed, "Time is so tight!"

"We need to repeatedly collect samples, collect pictures from various angles, and plan various movements. We took more than 200 photos in total for learning, and repeated the training 50 times. During training, we used the built-in GPU of Jeston to finally complete the walking of the self-driving car. In the end, four data are closely connected, "overall speed", "steering acceleration", "differential processing of walking trajectory", and "balance of left and right motors". It is really not easy to complete a simple track by combining "environment", "site", "data", and "four major variables"! ” ——–Chen, a student from Zhengxin High School

Competition Champion-"It's not as easy as you think after you try it!"

"Our group's Jetbot had some problems, so we started testing it relatively late. We also encountered many problems during the production process, such as being unable to connect to the computer or finding the robot on the computer. In the afternoon, when we were operating the robot and taking photos for training the model, we encountered program failures or errors one after another, and the robot ran out of power without warning. The only solution was for us to slowly find the cause of the problem and troubleshoot it, and find the appropriate charger to charge the robot. " --- Wu, a student from Xinmin High School

Jetbot self-driving car video recording

Not only are there practical courses, but the basic knowledge is also not neglected!

The first day of the four-day camp will lead everyone into the field of artificial intelligence, using courses on neural networks, machine learning, the Internet of Things, and Python programming language to allow students to have a deeper understanding of this field. The second day was an advanced understanding of the world of robotics, including embedded single-board computers (Jetson Family), robot brain and visual installation (OS & Camera), deep learning (NVIDIA Jetson-inference) and computer vision applications (detection, recognition, segmentation).

" On the second day, we used ThinkerCAD to simulate Arduino. In the afternoon, we first introduced edge devices, as well as the applications and architecture of Jetson in life. Then, we experienced a robotic arm and competed to use the robotic arm to clamp the yellow duck. Whoever clamped the most would win. After that, we assembled JetBot together. Fortunately, the power distribution inside is not difficult and can be understood with a little knowledge. In the course, I learned more about Arduino components and some theoretical knowledge. Although I have been exposed to some courses in school, I still listened to them with great interest. " - Wei, a student from Mingdao High School

Study resume, fully feedback the learning process and gains!

At the end of each day of the camp, students are given ten minutes to review the key points of today’s course and the knowledge they have absorbed, and to write a learning resume to have a complete record of their learning process. In addition, this camp also provides hands-on teaching to guide each student to obtain the NVDIA "Getting Started with AI on Jeston" international certification, enriching their own four-day learning files!

Four-day camp opens up high school students’ AI horizons!

This four-day course really brought many different experiences and brainstorming to each student. We open the door to the field of AI robots and self-driving cars for students, and hope that every student will be able to explore more deeply and create more self-driving car-related applications in the future!

"Four days ago, I had never come into contact with robots or Jetbots, but with the help of the teaching assistant and the teacher, I was able to complete the task successfully. I felt very accomplished and learned more knowledge that I didn't know before. I also had the opportunity to assemble a Jetbot, which was an unprecedented experience. If I have the opportunity to come into contact with this aspect again in the future, it will probably arouse my interest. " - Student Hong from Nanhu High School

Want to learn more? Click the button for more information!