Titulo Estágio
Deep Learning for Visual Robot Navigation
Áreas de especialidade
Sistemas Inteligentes
Local do Estágio
Departamento de Engenharia Electrotécnica da Universidade de Coimbra
Enquadramento
Mobile robots have been becoming more autonomous, but some challenges persist. One of the main challenges for a robot that aims to coexist with humans, is the ability to understand its environment and safely navigate through it. These skills are not specific for indoor robots, but also applicable for autonomous driving. Traditionally, the navigation relies on cloud-points from proximity sensors (such as Laser Range Finders). Humans and animals, on the other hand, use vision and their main source of information. For that reason, some work has already been done on using visual information for navigation [1,2]. Moreover, vision also enables the detection and classification of objects and places of interest, which aid in understanding the surroundings [3].1 - Lategahn, H., Geiger, A., & Kitt, B. (2011, May). Visual SLAM for autonomous ground vehicles. In 2011 IEEE International Conference on Robotics and Automation (pp. 1732-1737). IEEE.2 - Williams, B., & Reid, I. (2010, May). On combining visual SLAM and visual odometry. In 2010 IEEE International Conference on Robotics and Automation (pp. 3494-3500). IEEE.3 - Chen, C., Fragonara, L. Z., & Tsourdos, A. (2021). RoIFusion: 3D Object Detection From LiDAR and Vision. IEEE Access, 9, 51710-51721.
Objetivo
The main goal of this dissertation is to design, implement and test a framework that enables a robot to safely navigate an environment and identify places and objects of interest using visual information. The work will investigate the applicability of Evolutionary Machine Learning [4] techniques for producing Convolutional Neural Networks [5] which will perform this task.The various modules of the robot shall be integrated using the Robot Operating System (ROS) framework.4 - Mirjalili, S., Faris, H., & Aljarah, I. (2019). Evolutionary machine learning techniques. Springer-Verlag GmbH.5 - Bengio, Y. (2009). Learning deep architectures for AI. Foundations and trends® in Machine Learning, 2(1), 1-127.
Plano de Trabalhos - Semestre 1
1 - Literature review.2 - First steps with a mobile robot.3 - Definition of the techniques and technologies that will be used. 4 - System Architecture Design5 - Implementation of the first version of the system 6 - Writing the intermediate report
Plano de Trabalhos - Semestre 2
7 - Analysis of the first prototype and the obtained results8 - Refinement of the prototype9 - Validation in a realistic environment 10 - Writing the thesis11 - Writing a scientific article with the main results
Condições
The work will be carried out in the Laboratory of Embedded Systems of the Institute of System and Robotics (next to the Department of Electrical Engineering of the University of Coimbra).
Orientador
Lino Marques
lino@isr.uc.pt 📩