AutonomousEV3Car

Demo applications of autonomous, or self-driving, car by LEGO Mindstorms EV3.

If you wanted to see our car running, please watch the video on the YouTube.

We also have our website.

As you can see on the video, we have implemented the autonomous car as a internal research project.

Our car only uses an image (176*176, grayscale jpeg format) which is sent from iPhone set in front of the car.

This repo includes following demo applications.

• EV3Control: iOS App of remote controller of EV3 car
• ev3rt-utagoe: C++ program which controls EV3 directly
• inverse_reinforcement_learning: Demo application of Inverse Reinforcement Learning (IRL) with simple simulator using OpenGL
• lane_detection: Detecting lane using OpenCV
• lane_tracking: Tracking lane using Convolutional Neural Network (CNN)

For more informations, please check the README.md on each directory.

Getting Started

1. Building EV3 car
   We used a RAC3 Truck and remodeled a bit as follows.

   1. Put away the steering wheel.
   2. Add a vertical bar in the front of the truck to keep the iPhone in the right place.

2. Prepare SD card for inserting into the EV3
   There are two ways for preparing the SD card. The easiest way is to copy ev3rt-utagoe/sdcard/* on the SD card.

3. Run any application you like!

References

1. Deeper Depth Prediction with Fully Convolutional Residual Networks
2. iro-cp/FCRN-DepthPrediction
3. OpenCV Python Tutorial - Find Lanes for Self-Driving Cars (Computer Vision Basics Tutorial)
4. RyosukeHonda/Advanced_Lane_Lines
5. Canny法によるエッジ検出