Indoor localization of human workers in a robotized warehouse
Status: Looking for candidates
Description: Due to safety requirements there is need to develop algorithms for human localization in robotized warehouses (think of Amazon Robotics, i.e., Kiva warehouses). The idea is to develop an algorithm that will handle the 3 DoF localization problem only with a stereo camera and and an IMU worn by a human. The method will also use ID tag identification throughout the warehouse in order to minimize possibility of localization errors due to challenging conditions within the robotized warehouse. Main challenges arise from the fact that the localization is based on a fusion of sensor information and that all algorithms have to work in warehouse conditions in real time on onboard PC that can easily be worn by human without causing discomfort.
Contact: Asst. Prof. Ivan Marković, Prof. Ivan Petrović
Thesis type: Bachelor thesis, Master thesis, Semester project
High resolution head localization for augmented reality in robotized warehouses
Status: Looking for candidates
Description: The head orientation needs to be constantly computed in order to properly display all the information about the robots, objects and paths trough the warehouse using augmented reality glasses (e.g., Google Glass, Microsoft Hololens). This can be achieved by combining image features with the data from IMU sensors. The goal of this project is to develop algorithms for head orientation estimation based on sensors in the augmented reality glasses. Possible applications are not limited just to augmenting reality for workers in warehouses, but also to any application which requires accurate embedding of images depending on the structure of the environment.
Contact: Asst. Prof. Ivan Marković, Prof. Ivan Petrović
Thesis type: Bachelor thesis, Master thesis, Semester project
Human intention and plan recognition based on the Bayesian theory of mind
Status: Closed
Description: Human brain is the most sophisticaed recognition system we know. The theory of mind is our intuitive conception of other people's mental states - beliefs and desires - and how they cause bahviour. The goal of this thesis is to recognize human intentions and plans based on belief and desire depended planning and reconstruct the agent's belief state using Bayesian inference. The applications of this approach are numerous, but the thesis will focus on two problems: 1) recognizing workers intention of not following a predefined trajectory and 2) recognizing worker's intentions to leave a working station and alert the worker in case a task has remained unfinished.
Contact: Asst. Prof. Marija Seder, Asst. Prof. Ivan Marković, Prof. Ivan Petrović
Thesis type: Master thesis
Omnidirectional vision for autonomous mobile robots
Status: Looking for candidates
Description: Omnidirectional cameras by their definition provide a 360 degree view of the surrounding scene and as such pose themselves as a powerful tool in robot’s vision system (think of Google Street View). The enhanced field of view can be obtained by using several synchronized panoramic cameras, a combination of a camera and a mirror, or a camera with a wide-angle lens. The amount of information in such a single image reinforces robot’s abilities in interpreting and adequately acting and reacting in the environment. Besides robotics, these cameras can also be exploited in autonomous vehicles and surveillance systems. In this project we will focus on a standard perspective camera equipped with a fish-eye lens. The goal of the project is to develop novel tools based on omnidirectional cameras for autonomous mobile robot navigation.
Contact: Asst. Prof. Ivan Marković
Thesis type: Bachelor thesis, Master thesis, Semester project
Trajectory simulation of resident space objects
Status: Looking for candidates
Description: A Resident Space Object is a natural or artificial object that orbits another body, e.g., Sun, Earth, Mars etc. Here we are focusing on artificial objects that are Earth orbiting. As the space domain becomes increasingly congested due to global use of space and on-orbit collisions, there is a critical need to help better characterize this vital environment. The goal of this project is to develop trajectory simulation algorithms for resident space objects, i.e., in our case satellites, in near-geosynchronous orbit for the purposes of developing advanced tracking algorithms. These methods could then help to better track satellites and predict their future motion in order to avoid collisions and reduce the ever increasing problem of space pollution. The simulations will take into account as if measurement were taken from a true telescope, e.g., the optical telescope at Maui, Hawaii.
Contact: Asst. Prof. Ivan Marković
Thesis type: Bachelor thesis, Master thesis, Semester project
Detection of moving objects with the 3D laser range sensor mounted on a mobile robot
Status: Closed
Description: Detection and tracking of moving objects is an essential problem in situational awareness context and hence crucial for many robotic applications. The goal of this thesis is to develop a method for the detection of dynamic objects with a 3D laser range sensor and a variation of the method for tracking multiple detected objects. The ego-motion of the mobile platform should be compensated using the available odometry information and then refined using the point cloud registration techniques. It is expected that experiments will be performed using the Velodyne HDL-32E laser sensor mounted on top of a mobile platform.
Contact: Asst. Prof. Ivan Marković
Thesis type: Master thesis