Area III - Mobile Robotics


Nowadays, people are paying more and more attention to mobile robots due to their expectation of bringing huge impact on our lives, indutstry, and economy. Although indutrial robots have been used for decades, mobile robots will open completely new era with various application areas expanded by mobility and artifiical intelligence. The mobile robot team in CI lab, studies essential technologies for mobile robots, for instance, localization. Our research is focused on the field of robotics such as localization, map building, path planning, and 3D reconstruction.



Simulataneous Localization and Mapping


Simultaneous Localization and Mapping (SLAM) is an essential technique for a mobile robot to perform its missions in an unknown environment. SLAM merges uncertain meansurements from sensors to estimate the state of the robotí»s pose and the map of environment. The main problems consist of feataure extraction, data association, and estimation, each of which is our research topic. We are mainly studying Ceiling Vision (CV) SLAM which uses a camera looking up a ceiling for indoor environment perception and Exteded Kalman filter for estimation of the state. Related Video

-The map and trajectory estimated by SLAM-

-The mobile robot
platform
with a camera
and a laptop-

Also, we studied 3D SLAM using RGB-D sensor and a robot. As the robot goes along the hallway, it updates probability of all the hypothesis using sensor data and estimate its location using that hypothesis.Related Video

- 3D map estimated and reconstructed by SLAM -



Autonomous Driving


Autonomous Navigation is a comprehensive technique that includes localization, path planning, collision avoidance and so on. Localization is achieved by the particle filter with the map of environment. Local path planning and collision avoidance problems are solved by the Dynamic Window Approach (DWA) for safe traveling. The A* algorithm is used for global path planning which plans a long term trajectory. The various algorithms are needed to implement the autonomous navigation system, which enables a robot to carry out the missions. Related Video

- Vision Marker based Autonomous Driving -



Global Localization


Global localization means finding out the position of the robot with no priori information of the position in the environment map. The robot localizes itself by matching sensor measurements with the map. To solve this problem, we use Monte Carlo Localization (MCL) based on random sampling or Multiple Hypothesis Tracking methods. Related Video


- Example of Global Localization -



3D reconstruction


3D reconstruction gathers real-world information and fuse them into a single 3D model so it can be used in vast applications such as SLAM, gaming and even in medical science. With the development of depth sensors, it is possible to reconstruct detailed 3D model representing real-world online. The main focus of our research is algorithm called KinectFusion. A depth frame acquired by a depth sensor is fused incrementally into a single 3D reconstruction using the volumetric TSDF(Truncated signed distance function) and at the same time, it tracks the movement of the camera simultaneously. Related Video



- Example of 3D reconstruction -

   
 

Copyright(c) 2006. Computational Intelligence Lab.
All rights Reserved.
TEL: +82-2123-2863 / E-mail: etkim@yonsei.ac.kr

 
Send an E-mail Go E3 Go Yonsei