TatamiRacer: Development of a Tiny Autonomous Car for Education and Hobby
Introduction
In recent years, autonomous driving technology has developed rapidly in the industrial sector, prompting accelerated real-world implementation. At the same time, programming education and STEM (Science, Technology, Engineering, Mathematics) are gaining importance in educational settings, driving the demand for hands-on teaching materials. In this study, we present “TatamiRacer,” a miniature autonomous car built on a Tamiya Mini 4WD chassis, designed to safely demonstrate and teach autonomous driving concepts in a limited space. TatamiRacer combines a Raspberry Pi, a motor driver, a camera module, and the Donkey Car software framework.
The main features of TatamiRacer include:
Low Cost and Space Efficiency: The entire system operates in an area the size of a single tatami mat (approximately 1.8m x 0.9m) and uses inexpensive components.
Educational Design: It can serve as a teaching tool for students ranging from upper elementary school to university, covering programming and the fundamentals of deep learning.
Extensibility: Centered on a Raspberry Pi, it is easy to integrate additional sensors or networking features as needed.
This paper provides a comprehensive overview of the system’s hardware configuration, software installation, assembly procedures, and the learning process involved.
Related Works
Numerous efforts have been reported worldwide to replicate autonomous driving on small platforms. For instance, many educational materials rely on RC cars equipped with Raspberry Pi or Jetson Nano to perform line tracking or object recognition using OpenCV or TensorFlow. Below, we summarize some key points in existing projects to highlight the unique aspects of TatamiRacer.
Educational Use of Small Autonomous Platforms
Donkey Car: An open-source autonomous driving platform used in this study. It is Python-based, highly extensible, and supported by a large user community. Typically, Donkey Car runs on 1/10-scale RC cars that require a spacious area for operation, making it less convenient when space is limited.
Arduino-Based Mini Cars: Many examples feature Arduino Uno combined with ultrasonic or infrared sensors for line tracking or obstacle avoidance in small cars. This approach is beginner-friendly but limited by hardware constraints, making it less suitable for deep learning tasks or camera-based processing.
Deep Learning and Autonomous Driving
TensorFlow and PyTorch for Object Detection: Experiments are conducted in educational settings where cameras are mounted on small cars to facilitate object detection and control. However, setting up a deep learning framework from scratch can be time-consuming, and budget or time constraints often limit the depth of these projects.
Convenience of Donkey Car: The Donkey Car framework streamlines the entire pipeline—from data collection to training and inference—abstracting many aspects of vehicle control. This makes it feasible for educational environments and individual hobbyists to develop prototypes of autonomous cars powered by deep learning.
Use of Tamiya Mini 4WD
Mini 4WD as a Culture of Modification: In Japan, Tamiya Mini 4WD is widely recognized as a hobby culture, with enthusiasts modifying components such as the axle, gear ratio, body design, and weight. In educational contexts, these modifications provide practical examples for learning mechanical concepts such as gears, motors, and chassis design.
Positioning of TatamiRacer
Based on this context, TatamiRacer stands out for the following reasons:
Space-Saving Operation: A single tatami-sized area (about 1.8m x 0.9m) is sufficient for testing and demonstrations, making it practical for typical homes or classrooms.
Modular Structure: Raspberry Pi, motor drivers, cameras, and other off-the-shelf components can be easily procured and replaced.
Integration of Learning and Practice: The Donkey Car framework allows users to study the processes of deep learning and inference while conducting real-world driving experiments.
This unique set of attributes makes TatamiRacer an appealing educational and research tool compared to other small-scale autonomous driving platforms.
System Overview and Implementation Steps
This section outlines the hardware and software components that make up TatamiRacer, as well as assembly and calibration procedures. Key points are summarized below.
1. Hardware Configuration
Base Chassis: A Tamiya Mini 4WD VZ chassis is used.
3D Parts: Specially designed 3D-printed parts enable secure mounting of the Raspberry Pi and camera module onto the Mini 4WD chassis.
Control Unit: Raspberry Pi Zero 2 W or Raspberry Pi 4 can be selected based on performance requirements.
Camera: A camera module (OV5647 or IMX219) captures real-time road images for inference.
Power Supply: A 5V/2.1A mobile battery with attention to size and weight for easy mounting.
Drive System: A DC motor (low RPM 130 type) and a servo (SG90) provide rear-wheel drive and front-wheel steering, respectively.
Motor Driver: An H-bridge module (e.g., L298N) for controlling motor rotation in both forward and reverse directions.
2. Software Installation
Donkey Car Installation: After the initial Raspberry Pi OS setup, install Python and clone the Donkey Car repository.
Virtual Environment Setup: Use venv or conda to create a dedicated environment and install all necessary dependencies.
Calibration: Adjust the steering angle and motor speed so that the vehicle drives straight.
Data Collection: Perform manual driving while recording camera footage, steering angles, and throttle values.
Deep Learning Model Training: Use the collected data to train the Donkey Car CNN model, predicting throttle and steering commands.
Autonomous Driving Test: Deploy the trained model to the car, and evaluate its performance on a designated track.
Potential Applications in Education
TatamiRacer can be leveraged in a range of educational scenarios for programming and STEM education. Below are some specific examples:
Programming Fundamentals
For upper elementary to junior high school students, only a subset of Donkey Car’s features may be needed. Even basic code modifications or parameter adjustments in Python teach important concepts such as sensor data usage and feedback control.
Practical Deep Learning
Students can deepen their understanding of the CNN training process, exploring how changes in the quantity of training data or network architecture affect driving accuracy and stability. They can also analyze misclassification cases and learn hyperparameter tuning, acquiring hands-on deep learning know-how.
Creative Project-Based Learning
Because the system involves designing or customizing 3D-printed parts, as well as modifying the Mini 4WD body, learners gain experience from an engineering design perspective. Moreover, it is possible to create a miniature course that resembles a tatami mat, making it ideal for interactive workshops and hands-on events.
Events and Exhibitions
Demonstrating TatamiRacer at school festivals or company exhibitions can spark interest in robotics and AI among a wide audience, showcasing the excitement and accessibility of modern technology.
Experimental Results and Discussion
In this project, we constructed a track approximately the size of a tatami mat (about 1.8m x 0.9m) and conducted multiple laps to test the car’s performance. By using a 5:1 gear ratio and a low-speed 130 motor, the vehicle could drive at a slower yet more stable speed. Multiple manual driving sessions were recorded to obtain diverse training data, and once the CNN model was trained, the car generally stayed on track without deviation.
Conclusion
This paper presented TatamiRacer, a miniature autonomous car based on Tamiya Mini 4WD, along with specific methods for educational deployment. TatamiRacer features a space-saving design that allows autonomous driving on a single tatami mat, coupled with flexible deep learning control provided by the Donkey Car framework. By using this platform, a wide range of hobbyists and learners can experience a hands-on educational environment spanning introductory programming to advanced deep learning applications.
Future challenges include improving driving accuracy under various conditions, expanding learning materials, and fostering collaborative learning initiatives. Addressing these challenges will enable TatamiRacer to make an even greater contribution to both education and research.
