Assigned Student Projects

Spring 2018

  • Project 1: Localization in smoke-filled environments through ultrasound sensing
Description: Aerial robots are required to operate in challenging environments that may be visually-degraded. Indicative examples are those of operation above or through smoke in case of a fire, or the operation inside city tunnels in case of accidents, natural disasters and other situations that require search-and-rescue response. Especially when the relevant environment is GPS-denied (e.g., a tunnel), the vehicle must be able to estimate its position and orientation despite the visually-degraded conditions. For this reason, in this project the goal is to investigate the potential of using ulstrasound sensors for the purposes of localization in GPS-denied and smoke-filled rooms. When an environment is with good light conditions, clear from obscurants, and geometrically featureful, camera- and LiDAR-based solutions work reliably which however is not the case when smoke is present. Motivated by nature (i.e., bats flying inside a cave), the ultrasound modality can offer a useful alternative to solve the robot localization problem.

  • Task 1: Literature review
  • Task 2: Proposition of Hardware System and Research Direction
  • Task 3: Hardware prototype implementation
  • Task 4: Localization algorithm development
  • Task 5: Method and system verification
  • Task 6: Extensive evaluation and report

Initial literature elements:
  • Drumheller, Michael. "Mobile robot localization using sonar." IEEE transactions on pattern analysis and machine intelligence 2 (1987): 325-332.
  • Brown, Russell G., and Bruce Randall Donald. "Mobile robot self-localization without explicit landmarks." Algorithmica 26, no. 3-4 (2000): 515-559.
  • Varveropoulos, Vassilis. "Robot localization and map construction using sonar data." The Rossum Project 10 (2005): 1-10.

Download camera data.

Project 2: Target/object-tracking through computer vision
Description: The ability of detecting an object (or target) of interest and tracking its motion is fundamental for multiple aerial robotic applications. In this project the goal is to investigate the potential of computer vision methods that employ a single camera to track a moving target onboard a moving platform (the aerial robot). The required result of the project is an accurate derivation of the centroid of a certain user-defined object in the camera frame, while the object itself is moving. Furthermore, the algorithm should be able to deal with phases of short-term occlusion, predict the trajectory of the object and re-detect once it appears again in the camera frame. 

  • Task 1: Literature review
  • Task 2: Proposition of Hardware System and Research Direction
  • Task 3: Camera integration in ROS and development of testing case (tracked object etc)
  • Task 4: Object tracking algorithm development
  • Task 5: Method verification
  • Task 6: Extensive evaluation and report

Initial literature elements:
  • Baker, Simon, Ralph Gross, and Iain Matthews. "Lucas-Kanade 20 years on: a unifying framework." (2003).
  • Tomasi, Carlo, and Takeo Kanade. "Detection and tracking of point features." (1991).
  • Yilmaz, Alper, Omar Javed, and Mubarak Shah. "Object tracking: A survey." Acm computing surveys (CSUR) 38, no. 4 (2006): 13.

Download the data. 

Project 3: Human detection through IR Vision
Description: In many search and rescue applications, the task is to identify where humans are and to register them on a map. In this process, the task of human detection is particularly important. Currently, deep learning algorithms present excellent performance in detecting humans on visible camera frames when the posture of the human is similar to that of the training data (e.g., a standing pedestrian). However, this accuracy drops suddenly when the human posture is different (e.g., lying down) or the person is within a very cluttered scene (e.g., a victim of an earthquake within the remainings of a collapsed building). A natural alternative is to employ thermal vision. In this project, the goal is to develop efficient computer vision algorithms that can enable human detection from lightweight low-resolution IR cameras. 

  • Task 1: Literature review
  • Task 2: Understanding IR camera data
  • Task 3: Human Detection on IR frames algorithm development
  • Task 4: Method verification
  • Task 5: Extensive evaluation and report

Initial literature elements:
  • Portmann, Jan, Simon Lynen, Margarita Chli, and Roland Siegwart. "People detection and tracking from aerial thermal views." In Robotics and Automation (ICRA), 2014 IEEE International Conference on, pp. 1794-1800. IEEE, 2014.
  • Davis, James W., and Mark A. Keck. "A two-stage template approach to person detection in thermal imagery." In Application of Computer Vision, 2005. WACV/MOTIONS'05 Volume 1. Seventh IEEE Workshops on, vol. 1, pp. 364-369. IEEE, 2005.
  • Fernández-Caballero, Antonio, José Carlos Castillo, Javier Martínez-Cantos, and Rafael Martínez-Tomás. "Optical flow or image subtraction in human detection from infrared camera on mobile robot." Robotics and Autonomous Systems 58, no. 12 (2010): 1273-1281.

Project 4: Aggressive maneuver trajectory generation
Description: The ability of an aerial vehicle to fly with agility depends simultaneously on its design, the implement flight control laws and the generated trajectory of reference. For the case of a quadrotor aerial robot, a trajectory characterized by often need for sharp 90-degree turns cannot be tracked with high speed. However, exploiting our model knowledge for the vehicle dynamics we can develop optimal trajectory generation algorithms that derive optimized paths that allow a Micro Aerial Vehicle (such as the quadrotor) to follow a sequence of viewpoints with high speed and agility. The key property in this process of the “differential flatness” of the system. This project takes place primarily in simulation, while experimental realization is a plus. 

  • Task 1: Literature review
  • Task 2: Model of the quadrotor MAV for trajectory generation purposes
  • Task 3: Trajectory generation algorithm development
  • Task 4: Method verification
  • Task 5: Extensive evaluation using the RotorS simulator and report

Initial literature elements:
  • Mellinger, Daniel, and Vijay Kumar. "Minimum snap trajectory generation and control for quadrotors." In Robotics and Automation (ICRA), 2011 IEEE International Conference on, pp. 2520-2525. IEEE, 2011.
  • Ritz, Robin, Mark W. Müller, Markus Hehn, and Raffaello D'Andrea. "Cooperative quadrocopter ball throwing and catching." In Intelligent Robots and Systems (IROS), 2012 IEEE/RSJ International Conference on, pp. 4972-4978. IEEE, 2012.
  • Müller, Mark, Sergei Lupashin, and Raffaello D'Andrea. "Quadrocopter ball juggling." In Intelligent Robots and Systems (IROS), 2011 IEEE/RSJ International Conference on, pp. 5113-5120. IEEE, 2011.

Project 5: Develop a passive legs-based mechanism for landing on rough terrain
Description: Aerial robots are often used in areas where take-off and landing may have to take place over anomalous terrain. Traditionally, people employ some sort of ad-hoc landing base if the terrain is so rough that it is considered it may hinder the safety of the vehicle. This contradicts the intuition about how seamlessly flying creatures deal with such situations. The key is that unlike most Vertical Take Off and Landing Micro Aerial Vehicles that employ fixed and rigid landing mechanisms, both bugs and bird have legs. In this project we consider the design of a passive mechanism that can enable safe landing above anomalous terrain and therefore provide inherent safety into the vehicle. The mechanism must be lightweight, able to be retracted during flight, and robust to forcible collisions. 

  • Task 1: Literature review
  • Task 2: Conceptual designs
  • Task 3: Design selection and manufacturing
  • Task 4: System verification
  • Task 5: Extensive evaluation and report


Initial literature elements:

  • Raibert, Marc H. "Legged robots." Communications of the ACM 29, no. 6 (1986): 499-514.
  • Hutter, Marco, C. David Remy, and Roland Siegwart. "Design of an articulated robotic leg with nonlinear series elastic actuation." In Mobile Robotics: Solutions and Challenges, pp. 645-652. 2010.
  • Delcomyn, Fred, and Mark E. Nelson. "Architectures for a biomimetic hexapod robot." Robotics and Autonomous Systems 30, no. 1-2 (2000): 5-15.

Presentation of your Project:
Each team should present their progress presentation in class on April 11. Please use this template and prepare ~10 slides for your project. Select one or more of your colleagues and team members to do the presentation. 

Spring 2016

Team 1: Development of an automated Fixed-Wing UAV [Download presentation] 

Team 2: Development of a Multirotor Aerial Vehicle capable of Autonomous Navigation [Download presentation] 

Team 3: Monocular, hardware-synchronized, Camera-IMU Localization [Download presentation] 

Team 4: Monocular, software-synchronized, Camera-IMU Localization [Download presentation] 

Team 5: Aerial Robot Swarming subject to Communication Constraints [Download presentation] 

Team 6: Hardware-in-the-Loop Fixed-Wing Control [Download presentation] 

Team 7: Large baseline Stereo Camera Depth Sensing [Download presentation] 

Team 8: Agile Multirotor Flight and Control through Smart Devices [Download presentation] 

Team 9: Simulation-based Control of Multirotor Aerial Vehicles [Download presentation]