Projects for Master Thesis and/or Independent Study
Informative Path Planning for Subterranean Flying Robots Supervisors: Kostas Alexis, Christos Papachristos Available This project aims to investigate the development of path planning strategies optimized for the case of underground environments exploration. We look for a path planning strategy that presents three behaviors in relation to if the setting is a) like a man-made tunnel (e.g., an underground mine), b) a multi-storey urban facility (e.g., a metro subway) or c) a natural cave. The methods must be computationally efficient and exploit the history of observations of the robot. Due to the complexity and multi-branching geometry of underground environments, a critical aspect of the solution is that of understanding of exploration completeness.
Aerial on Ground Robot Ferry for Underground Co-Deployment Supervisors: Kostas Alexis, Christos Papachristos Available This project investigates the potential of ground-and-aerial robot collaboration for underground environment exploration through specifically emphasizing on the potential of aerial-on-ground robot ferry. As such, a ground vehicle characterized by superior performance should only deploy its flying companion when this navigation modality is crucial for the mission success and contributes the most it can given its limited flight time. We seek both the methods to decide the collaboration strategy and students involved in the relevant mechatronic design.
Collision Resilient Navigation for Aerial Robots in Confined Environments Supervisors: Kostas Alexis, Christos Papachristos Available This research aims to enhance the navigation abilities of aerial robots in confined/narrow environments by enabling them to stably establish contact with surfaces of their environment. More specifically, a Micro Aerial Vehicle will be enhanced with specialized mechanisms for physical interaction (extensions) and a software framework for control in confined spaces by exploiting contact. Force feedback at the end effectors will facilitate stable and sustainable physical interaction. This research direction plans to radically change how flying robots navigate through narrow environments such as ore-passes or manholes.
Reactive Collision Avoidance for Aerial Robots Navigating in Underground Environments Supervisors: Kostas Alexis, Christos Papachristos Available This research aims to investigate a last-resort collision-avoidance mechanism that is implemented onboard an aerial robot aiming to navigate complex underground settings. The method should employ different sensing modalities in order to provide robustness and satisfactory performance even in visually-degraded conditions. In particular, the combination of visible camera data and LiDAR ranging is considered as a starting point. Ideally, the designed solution should not assume that a reliable pose and map estimate is available and enable reactive avoidance even in the most degenerate cases.
Simulation Environment for Subterranean Robotics Supervisors: Kostas Alexis, Christos Papachristos Available This project aims to develop a comprehensive simulation environment for testing ground and flying robots for autonomous navigation in subterranean environments. It will be based on Gazebo and ideally should also allow for some type of Hardware-In-the-Loop functionality. It will develop on top of the Gazebo-based simulator of subterranean environments provided by DARPA.
Object Detection and Classification in Thermal Vision Supervisors: Kostas Alexis, Christos Papachristos Available In a large variety of degraded visual environments, robotic classification of objects of interests becomes particularly hard due to the impaired sensor data. Especially in the case of darkness and obscurants, classical visible spectrum cameras and associated techniques are rendered ineffective. The goal of this project is to investigate the potential of using thermal vision camera data for the purposes of object detection and classification. Thermal vision is unaffected by darkness and can penetrate certain types of obscurants therefore offering a viable alternative. The envisioned research relates to the associated machine learning methods for high accuracy results on thermal vision. It involves both the steps of data annotation, alongside neural network design and training.
Leg-wheel Robot for Underground Navigation Supervisors: Kostas Alexis, Christos Papachristos Available In this project we consider the problem of developing a small and cost efficient robot for undergound navigation. A leg-wheel robotic design is proposed on the basis of the combined simplicity and ability to overcome rough terrain that these mechanisms present. The research and development tasks of the project relate to the mechatronic design, automated and control and autonomous navigation solution for such a robotic platform.
Understanding the Vision System of Underground Species Supervisors: Kostas Alexis, Christos Papachristos Available In this project we question what is special and specific to the vision system of specials (from animals to insects) living underground. Good examples include the wolf spider or beavers. The emphasis is on literature study on the specific domain and derivation of conclusions on how certain principles may apply to robotic vision systems both in the sense of hardware and algorithms. It corresponds to a project that will lay the ground for many subsequent investigations to follow.
Deep Learning for Monocular Thermal Vision to Depth Estimation Supervisors: Kostas Alexis, Christos Papachristos Available Lightweight depth perception is critical for underground navigation as it allows small systems (which can go through tight spaces) to proceed autonomously. Camera-based solutions are ideal but depth perception on monocular is ill-conditioned. In fact, visible-light cameras soon become degraded in the underground domain due to issues such as darkness and extreme dust. On the contrary, thermal vision penetrates darkness and obscurants. The goal of this project is to employ supervised deep learning to allow a pixel-wise depth value prediction from the thermal image given training in analogous environments.
Collaborative Visual-Inertial Localization and Mapping Supervisors: Kostas Alexis Assigned Robotic systems are not bound to operate alone. In this project the goal is to enable two robots to perform GPS-denied localization in a collaborative manner and simulatneously map their environment. Both robots will be equipped with a visual-inertial sensor. The problems of data association and establishing a common reference frame will therefore have to be addressed.
Pedestrian Detection and Avoidance Supervisors: Kostas Alexis Assigned In the very near future, cars are expected to be fully robotized and offer us back a critical subset of our time, while ensuring optimal safety. But for this goal to flourish, critical technologies have to be developed. Within this project, the goal is to enable fast and efficient pedestrian detection, intent recognition (prediction of expected trajectory) and subsequent planning of control actions for a car so as to avoid the pedestrian. A combination of visual and LiDAR data will be utilized for the optimal achievement of this goal.
Localization using a Solid-State LiDAR Supervisors: Kostas Alexis Assigned LiDAR technology has been particularly succesfull in enabling robots to localize themselves in GPS-denied environments. However, in most such cases the corresponding sensors are both expensive and of some significant weight. Nowadays, there is a lot of research towards the development of solid-state LiDAR systems. In this project, the goal is to develop a localization pipeline that will be efficient given the data provided by such a system, exploit its particular properties and deal with its limitations.
Localization and Mapping using a Thermal Camera Supervisors: Kostas Alexis Assigned Thermal imaging has certain properties that make it interesting as a sensing modality to assist the problem of localization and mapping for robitc systems. In many environments that good visual features may not be available, the thermal image may turn up being rich in information. However, certain challenges also exist due to the different charactecteristics of thermal imaging and how different viewpoints do not lead to the same differences in image information as in visible-spectrum imaging. The goal of this project is to develop a Simultaneous Localization and Mapping optimized for the case of thermal imaging onboard robotic systems.
Real-time Semantic Classification using onboard GPU Supervisors: Kostas Alexis Assigned. This project aims to develop a GPU-based architecture to enable the real-time, efficient and robust semantic classification of objects within the environment. The goal is to contribute a new approach that exploits the superior abilities of GPUs for parallel computing and eventually allow the real-time classification using both 2D and 3D information as provided from the sensors onboard a small aerial robot. To that end, an embedded GPU system is also employed.
Cooperative Autonomous Exploration using Aerial Robots Supervisors: Kostas Alexis (UNR), Mina Kamel (ETH Zurich) Available This project aims to develop cooperative and distributed algorithms capable for efficiently solving the problem of autonomous exploration based on a team of possibly heterogeneous aerial robots. Employing information gain–based approaches, we search to develop advanced methods that will allow a single aerial robot to explore its 3D environment autonomously and efficiently, while methods of direct/indirect communication will enable distributed cooperative exploration. The methods to be developed will be tested in a high–fidelity simulation environment that incorporates the sensor models and accounts for the true vehicle limitations and dynamics while the possibility of real–life experiments is provided. Download the Project Description (PDF)
Autonomous Driving on Snow and Ice Supervisors: Kostas Alexis Available For the autonomous cars to be able to satisfy their promise, they have to be able to deal with all environments. Among others, driving on snow and ice corresponds to a great challenge for the control, planning and perception loops of the robot. In this project, the goal is to utilize a small ground robot in order to develop the relevant control and planning algorithms and deliver a library for autonomous driving on snow and ice.
GPU-based Dense Mapping for Aerial Robotics Supervisors: Kostas Alexis (UNR) Assigned Dense mapping in real–time is among the most critical functionalities that an aerial robot should possess in order to be able to conduct applications useful to the society. Although some solutions exist, most of them rely on structured–light sensors and can operate only in very limited distances and appropriate light conditions. Using a stereo camera for this purpose can lead to a more versatile framework but achieving equal levels of data density in real–time and at comparable update rates is a major challenge. This project aims to examine the use of embedded systems with GPUs and stereo camera systems in order to achieve this goal. Download the Project Description (PDF)
Own ideas - high risk projects!
Do you have your own idea about a robotics project? Are you willing to discuss a high-risk project with the understanding that things might not always work? Contact me and schedule a meeting!