Revolutionizing Transportation: The Journey of Creating an Autopilot Bicycle

Table of Contents

1. Introduction
2. The Accident and Recovery
3. The Idea for an Autopilot Bicycle
4. Transforming the Bicycle
5. Developing the Control System
6. Designing the AI Computing Unit
7. Integrating the Components
8. testing and Fine-Tuning
9. Achieving Autopilot Capability
10. Future Improvements and Considerations

Introduction

In this article, we will explore the journey of creating an autopilot bicycle. We will delve into the inspiration behind this project, the challenges faced, and the steps taken to transform a regular bicycle into an autonomous vehicle. From designing the control system to integrating the AI computing unit, we will cover all aspects of this innovative venture. So, hop on and let's discover the world of self-driving bicycles!

The Accident and Recovery

It all started with an unfortunate accident. I fell while cycling and injured myself. During the recovery period, I had a lot of time to contemplate and came up with an interesting idea – to design a stabilizing device for bicycles. Coincidentally, the topic of autopilot technology was also gaining traction in the news. With my background in AI, I saw an opportunity to merge these two concepts and create something truly unique.

The Idea for an Autopilot Bicycle

To create an autopilot bicycle, I realized that I needed to tackle several fundamental challenges. Firstly, I needed to find a way for the bicycle to move on its own. Unlike a car with four wheels, a bicycle is less stable and requires a different approach. Secondly, I needed to eliminate the need for a human driver and equip the bicycle with sensors and a powerful computing unit to control its movements. Lastly, I needed to develop sophisticated Perception and control algorithms to enable the bicycle to make informed decisions.

Transforming the Bicycle

The first step towards achieving an autopilot bicycle was to transform a regular bicycle into a fixed gear (dead fly) bicycle. This type of bicycle has a simple structure without brakes and decelerates by reverse pedaling. By using computer-aided design (CAD) tools, I created a virtual model of the bicycle, allowing me to experiment with different modifications. The modifications included installing brushless motors, sensors, and a control system, as well as a depth camera and a laser radar.

Developing the Control System

Achieving balance was a crucial aspect of the control system. By studying the law of conservation of angular Momentum, I understood that the movement of the brushless motors could help maintain the balance of the bicycle. The brushless motors acted as a driving force, while the metal wheel behind them functioned similarly to a satellite adjusting its attitude. This innovative approach enabled the back and forth movement of the bicycle.

Designing the AI Computing Unit

The next step was to design the AI computing unit that would serve as the brain of the autopilot bicycle. I opted for a low-power, real-time operating system to control the body of the bicycle and a high-delay, non-real-time operating system for perception, decision-making, and thinking tasks. The integration of microcontrollers and AI chips allowed for efficient data processing, motor control, and sensor Fusion.

Integrating the Components

With the control system and AI computing unit in place, it was time to integrate all the components. The machining process involved a combination of 3D printing and metal processing to create the necessary parts. The assembly of the bicycle was relatively straightforward, as the design was based on my own drawings. However, the circuitry, including the main control module and AI computing unit, was not yet installed.

Testing and Fine-Tuning

Before achieving full autonomy, rigorous testing and fine-tuning were essential. The attitude control was based on an LQR controller and a classic PID controller, utilizing data from accelerometers and gyroscopes. The unity Game engine was used to create virtual simulations, allowing for the visualization and verification of the control algorithms. Sim2real results were obtained by migrating The Simulation to a realistic environment, accounting for uncertainties and enhancing the bicycle's performance.

Achieving Autopilot Capability

With the integration, testing, and fine-tuning complete, the autopilot bicycle was ready to hit the road. Equipped with a depth camera, the bicycle could perceive its surroundings and transmit real-time images to a computer. Combined with AI algorithms, the bicycle could perform tasks such as obstacle avoidance and object identification. The laser radar facilitated SLAM mapping and path planning, enabling the bicycle to explore its environment autonomously.

Future Improvements and Considerations

While the basic autopilot functions were successfully implemented, there is still room for improvement. For instance, the design of the leading structure could be optimized by incorporating harmonic gear direct-drive motors for enhanced stability. Additionally, adapting the system to electric bicycles could open up new possibilities. The open-source nature of the project allows for further exploration, and interested individuals can access the repository to replicate or enhance the design.

Highlights

• Creating an autopilot bicycle by merging the concepts of stability devices and autopilot technology.
• Transforming a regular bicycle into a fixed gear (dead fly) bicycle.
• Designing an AI computing unit as the brain of the autopilot bicycle.
• Integrating components such as brushless motors, sensors, and control systems.
• Testing, fine-tuning, and simulating the control algorithms for optimal performance.
• Achieving autopilot capabilities, including obstacle avoidance and path planning.
• Exploring possibilities for future improvements, such as harmonic gear direct-drive motors and electric bicycle adaptation.

FAQ

Q: Can the autopilot bicycle maintain balance even under sudden shocks or increased loads? A: Yes, thanks to the precise control algorithms and the integration of brushless motors, the autopilot bicycle can maintain balance even in challenging situations.

Q: Are there any plans to expand the capabilities of the autopilot bicycle in the future? A: Absolutely! The current version of the autopilot bicycle represents the basic implementation. There are plans to optimize the design, including the leading structure and the use of electric bicycles.

Q: How can I replicate or enhance the design of the autopilot bicycle? A: The entire project, including the structural design and circuitry, is open source. You can access the GitHub repository for detailed information and resources.

Q: What are some potential applications for an autopilot bicycle? A: An autopilot bicycle has the potential to revolutionize transportation, especially in urban areas with congested traffic. It can also be used for delivery services or as a recreational vehicle.

Q: Can the autopilot bicycle identify and follow objects in motion? A: Yes, with the combination of depth cameras, AI algorithms, and laser radar, the autopilot bicycle can detect and track various objects, including those in motion.

Resources

• GitHub Repository: [Link to GitHub Repository]
• Unity Game Engine: [Link to Unity Game Engine]
• Autonomous Vehicle Technology: [Link to Autonomous Vehicle Technology Website]