Obstacle Avoidance Robot Using Convolutional Neural Network

The aim of the project is to design an obstacle avoidance system by using neural network. The main functionality of the project is to predict the object detected and to find a path of avoidance the detected object. Due to the increase in autonomous vehicle centered around machine learning technology, expensive system which incorporates multiple sensors for the obstacle avoidance and object detection are made. To fill this gap, the project is designed to implement an object detection and obstacle avoidance system using a camera as the main component for detection. Jetson nano board is used as the main computer, NoirV2 camera is used as the main vision sensor, 3d printed structure for the body, two dc motor with a I2c based motor driver. SSD MobileNet v2 is used as the model for object detection, Jetson inference is the training guide which is used for the obstacle avoidance and object detection which is optimized to work with Jetson nano board providing a faster detection and fps. The project includes a web based Graphical user interface to control the robot and to monitor it. The project will solve the issues of expensive system and requirement of multiple sensors for object detection and obstacle avoidance.