Post date: Jun 6, 2012 9:28:44 AM
======================================================= Open position for PhD Student: Long-Term Visual Mapping ======================================================= Robotics, Perception and Real-Time Group at the I3A, Universidad de Zaragoza (Spain) is looking for PhD student candidates with excellent academic qualifications and interest in pursuing a PhD in the context of the RoboEarth European Project (http://www.roboearth.org/), devoted to create a knowledge base where robots can share world models and action recipes. How to apply ============ Please, send the following documentation to tardos.phd.position.12 at gmail.com: 1- C.V. 2- Motivation letter (career plan, motivation for PhD, statement of research interest) 3- Copy of original study certificates for your Degree and Master (if available), with list of subjects, grades, and scale (indicate min, max, and pass thresholds) 4- Copy of international publications, if available (max 3) 5- Two references, including its phone contact details Deadline for application: June 30th, 2012 e-mail: tardos.phd.position.12 at gmail.com Requirements: ============= Academic: - Excellent academic grades. - Degree on Computer Science, Electrical Engineering, Mechatronics, Maths, Physics or similar. - Master degree in these areas will be valuable but not imperative. Background: - Knowledge in robotics, computer vision and image processing. - Good programming skills in C++ and Matlab. - Previous experience in SLAM techniques will be valuable. Details of Position =================== Contract Type: Phd. Student Grant Start Date: September, 2012 Duration in months: 48 Salary: As determined by regulations of the Spanish PhD Student Grants (around 1200EUR/month) Place of Work: Universidad de Zaragoza, Zaragoza, Spain. Research Topic: LONG-TERM VISUAL MAPPING -------------------------------------------------------------------- This research work will consider the problem of long-term (or lifelong) visual mapping using monocular or stereo cameras. First, we will develop fast and robust place recognition techniques, using recently proposed binary features, specially the scale and rotation invariant ones. We will investigate which clustering or hashing techniques are more appropriate in binary spaces, and analyze their performance for loop closure and camera relocation using the bag-of-words technique. Second, we will address the problems of long-term mapping; detection of changes in the environment, map updating, removal of obsolete parts, and use of learning techniques to identify stable versus instable areas of the environment. The goal would be having a robot to autonomously learn that areas such as desktops are prone to change in appearance, and focus the relocation algorithms in the more stable regions.