Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/akshaymogaveera/self-driving-robot-using-neural-network

This project introduces the autonomous robot which is a scaled down version of actual self-driving vehicle and designed with the help of neural network. The main focus is on building autonomous robot and train it on a designed track with the help of neural network so that it can run autonomously without a controller or driver on that specific track. The robot will stream the video to laptop which will then take decisions and send the data to raspberry pi which will then control the robot using motor driver. This motor driver will move the robot in required directions. Neural Network is used to train the model by first driving the robot on the specially designed track by labeling the images with the directions to be taken. After the model is trained it can make accurate predictions by processing the images on computer. This approach is better than conventional method which is done by extracting specific feature from images.
https://github.com/akshaymogaveera/self-driving-robot-using-neural-network

arduino artificial-neural-networks driverless-car google gradient-descent keras machine-learning neural-network opencv opencv3-python python raspberry-pi robotics self-driving-car self-driving-car-simulation tensorflow tensorflow-models waymo

Last synced: 8 days ago
JSON representation

This project introduces the autonomous robot which is a scaled down version of actual self-driving vehicle and designed with the help of neural network. The main focus is on building autonomous robot and train it on a designed track with the help of neural network so that it can run autonomously without a controller or driver on that specific track. The robot will stream the video to laptop which will then take decisions and send the data to raspberry pi which will then control the robot using motor driver. This motor driver will move the robot in required directions. Neural Network is used to train the model by first driving the robot on the specially designed track by labeling the images with the directions to be taken. After the model is trained it can make accurate predictions by processing the images on computer. This approach is better than conventional method which is done by extracting specific feature from images.

Awesome Lists containing this project

README 

        
# Self-Driving-Robot-Using-Neural-Network 

### Python + Tensorflow + OpenCV

### Overview

   1) The Robot uses neural network to learn and predict decisions just like a human brain.

   2) The model is built using Neural Network and it's trained by feeding in labelled images of tracks.

   3) After the model is trained it will be capable of making its own decisions. The prediction will be made on the laptop due to larger memory and flexibility. Raspberry pi will be used to stream the video to laptop using Pi-camera.

   4) First we will train the model using the dataset that contains the labelled images of the track.

   5) Raspberry Pi will stream the live feed to the laptop and the predictions will be sent back to the raspberry pi.

   6) The raspberry pi is connected to motor driver which will control the wheels of the bot. Ultrasonic sensor makes sure that the robot does not collide with obstacles. Once trained it can run autonomously and make its decisions.It will try to maintain its path along the track and prevent from collisions. 



### For more detailed explanation, please view my IEEE paper titled Self Driving Robot using Neural Network or view PDF 

    

### Watch the Robot in Action !!

![alt text](https://github.com/akshay1997feb/Self-Driving-Robot-Using-Neural-Network/blob/master/pic.png)"



    

### A) Hardware Design

   The Hardware components used for this project are as follows:

    

   1. Raspberry pi-3.

   2. Pi Camera.

   3. Robot (Chassis, Wheels (4), Motors (2).

   4. Ultrasonic Sensor.

   5. Motor control (L293D).

    

   Raspberry pi will work as a brain of the robot, which takes all the decision of the robot and live streams video to the laptop. Ultrasonic sensor is used to calculate the distance of the obstacles ahead.

    









### B) Software Used:-

   1) Python(2.7)

   2) TensorFlow

   3) OpenCV



## Working

For detailed explanation about the model and working visit on the IEEE link given above.

* **Data Collection**

    * Images for training is collected by driving the robot.

    * Run File **"rpi_connection_cam.py"** on Raspberry Pi and simultaneouly run File **"collect_images_via_rpi.py"** on your PC. A server client connection is established between Rpi and your computer. Ensure that they are on the same Network.

    * Pygame is used for driving the robot, it is driven by using Keyboard (w-forward, a-left, d-right). When a key is pressed the image with its label is saved.

    * NumPy is used to save the data and its corresponding label.

* **Model Training**

    * Tensorflow is used to create the model.

    * After sufficient data is collected, run file **"train_neural_network.py"** on your PC (Ensure path where the model will be saved in code is correct).

    * Training the model will take some time and post completion the model will be saved.

* **Final**

    * Ensure that the previous steps are completed without any errors.

    * Run files **"rpi_connection_cam.py"** and **"rpi_controlcar_distance.py"** on raspberry pi simultaneously, parallely run file **"run.py"** on your PC.

 

 

 

 

# Credits @hamuchiwa, as I have followed his tutorial and used some of his code.