Study on electronics and communication engineering

Abstract

The main objective of the thesis is to learn policies which match the results from recorded data collected from agents in the real world, so that the vast volumes of the data in the real world can be made useful. The scopes of this project are verifying whether Deep Learning can be used successfully in Duckietown platform; By the concept of ‘Data Processing Inequality’, using supervised and imitation learning to control the Duckiebot end-to-end (input: compressed image, output: control command) with data from a recorded policy; Using supervised or unsupervised learning to model specific aspects of the autonomous driving task; Focus on autonomous lane following by learning based tools. Our team develop this project with step by step from research to implemetation. Firstly, we research relations and differences among machine learning, deep learning, imitation learning, supervised learning and unsupervised learning. Then, project will be trainned an effective Convolutional Neural Network which maps compressed image to orientation of bots for lane following (the most practical and difficult part). And our team implement a ROS node which subscribes to input images, communicates with Cloud Training Platform (Google Colab) for computation, and publishes the computed orientation angle to the car control node. Specifically, imitation learning for driving is a supervised learning based tool, that clones behavior. The experts can be humans or optimal/near optimal planners/controllers. In our project, we regard the conventional approach for lane following as optimal controllers and use it as expert to collect training data.