Theses in Robotics

From Intelligent Materials and Systems Lab

Projects in Advanced Robotics

The main objective of the follwing projects is to give students experience in working with advanced robotics tehcnology. Our group is active in several R&D projects involving human-robot collaboration, intuitive teleoperation of robots, and autonomous navigation of unmanned mobile platforms. Our main software platforms are Robot Operating System (ROS) for developing software for advanced robot systems and Gazebo for running realistic robotic simulations.

For further information, contact Karl Kruusamäe.

The following is not an exhaustive list of all available thesis/research topics.

Highlighted theses topics for 2023/2024 study year

  1. ROBOTONT: analysis of different options as on-board computers
  2. SemuBOT: multiple topics
  3. Quantitative Evaluation of the Situation Awareness during the Teleoperation of an Urban Vehicle
  4. ROBOTONT: ROS2 support for robotont
  5. Continuous teleoperation setup for controlling mobile robot on streets
  6. Stratos Explore Ultraleap demonstrator for robotics
  7. Mixed-reality scene creation for vehicle teleoperation

List of potential thesis topics

Our inventory includes but is not limited to:

ROBOTONT: analysis of different options as on-board computers

Currently ROBOTONT uses Intel NUC as an onboard computer. The goal of this thesis is to validate Robotont's software stack on alternative compute devices (e.g. Raspi4, Intel Compute Stick, and NVIDIA Jetson Nano) and benchmark their performance for the most popular Robotont use-case demos (e.g. webapp teleop, ar-tag steering, 2D mapping, and 3D mapping). The objetive is to propose at least two different compute solutions: one that optimizes cost and another that optimizes for the performance.


ROBOTONT: integrating a graphical programming interface

The goal of this thesis is to integrate graphical programming solution (e.g. Scratch or Blockly) to enable programming of Robotont by non-experts.


SemuBOT: multiple topics

In 2023/2024 many topics are developed to support the development of open-source humanoid robot project SemuBOT (https://www.facebook.com/semubotmtu/)


ROBOTONT: Quick launch demo suite and the final complete release of its ROS1 software

The goal of this thesis is to refine existing robotont demos (e.g. ar-tag steering, follow-the-leader, dancing-with-robot, LEAP-based control etc) and package them in an easy to use way for quick deployment by anyone during public events such as science popularization workshops and school visits. The results of this work will be packaged as the final ROS1 release of Robotont software as the EOL for ROS1 is in May 2025.


ROBOTONT: ROS support, demos, and educational materials for open-source mobile robot

ROS

The project involves many potential theses topic on open-source robot platform ROBOTONT. The nature of the thesis can be software development to improve the platform's capabilites, simulation of specific scenarios, and/or demonstration of ROBOTONT in real-life setting. A more detailed thesis topic will be outlined during in-person meeting

What is ROS?


Virtual reality user interface (VRUI) for intuitive teleoperation system

Detecting 2 hands with Leap Motion Controller

Enhancing the user-experience of a virtual reality UI developed by Georg Astok. Potentially adding virtual reality capability to a gesture- and natural-language-based robot teleoperation system.
Gesture-based teleoperation


Health monitor for intuitive telerobot

Intelligent status and error handling for an intuitive telerobotic system.


3D scanning of industrial objects

Using laser sensors and cameras to create accurate models of industrial products for quality control or further processing.


Modeling humans for human-robot interaction

True human-robot collaboration means that the robot must understand the actions, intention, and state of its human partner. This work invovlves using cameras and other human sensors for digitally representing and modelling humans. There are multiple stages for modeling: a) physical models of human kinematics and dynamics; b) higher level-models for recognizing human intent.
ROS & Kinect & Skeleton-Markers Package


Robotic avatar for telepresence

Integrating hand gestures and head movements to control a robot avatar in virtual reality user interface.


Detection of hardware and software resources for smart integration of robots

Vast majority of today’s robotic applications rely on hard-coded device and algorithm usage. This project focuses on developing a Resource Snooper software, that can detect addition or removal of resources for the benefit of dynamic reconfiguration of robotic systems. This project is developed as a subsystem of TeMoto.


Sonification of feedback during teleoperation of robots

Humans are used to receiving auditory feedback in their everyday lives. It helps us make decision and be aware of potential dangers. Telerobotic interfaces can deploy the same idea to improve the Situational Awareness and robotic task efficiency. The thesis project involves a study about different sonification solutions and implementation of it in a telerobotic application using ROS.


Human-Robot and Robot-Robot collaboration applications

Creating a demo or analysis of software capabilities related to human-robot or tobot-robot teams

  • human-robot collaborative assembly
  • distributed mapping; analysis and demo of existing ROS (e.g., segmap https://youtu.be/JJhEkIA1xSE) packages for multi-robot mapping
  • Inaccessible region teamwork
    • youbot+drone - a drone maps the environment (for example a maze) and ground vehicle uses this information to traverse the maze
    • youbot+robotont - youbot cannot go up ledges but it can lift smaller robot, such as robotont, up a ledge.

Mirroring human hand movements on industrial robots

The goal of this project is to integrate continuous control of industrial robot manipulator with a gestural telerobotics interface. The recommended tools for this thesis project are Leap Motion Controller or a standard web camera, Ultraleap, Universal Robot UR5 manipulator, and ROS.


ROBOTONT: TeMoto for robotont

Swarm-management for robotont using TeMoto framework.


Enhancing teleoperation control interface with augmented cues to provoke caution

The task is to create a telerobot control interface where video feed from the remote site and/or a mixed-reality scene is augmented with visual cues to provoke caution in human operator.


Robot-to-human interaction

As robots and autonomous machines start sharing the same space a humans, their actions need to be understood by the people occupying the same space. For instance, a human worker needs to understand what the robot partner is planning next or a pedestrian needs to clearly comprehend the behaviour of a driverless vehicle. To reduce the ambiguity, the robot needs mechanisms to convey its intent (whatever it is going to do next). The aim of the thesis is to outline existing methods for machines to convey their intent and develop a unified model interface for expressing that intent.


Gaze-based handover prediction

When human needs to pass an object to a robot manipulator, the robot must understand where in 3D space the object handover occurs and then plan an appropriate motion. Human gaze can be used as the input for predicting which object to track. This thesis activities involve camera-based eye tracking and safe motion-planning.


Real-world demonstrator for MIR+UR+TeMoto integration

Integration of mobile manipulator (MIR100 + UR5e + Robotiq gripper) to demonstrate TeMoto in a collaborative application.


ROBOTONT: Human-height human-robot interface for Robotont ground robot

Robotont is an ankle-high flat mobile robot. For humans to interact with Robotont, there is a need for a compact and lightweight mechanical structure that is tall enough for comfortable human-robot interaction. The objective of the thesis is to develop the mechanical designs and build prototypes that ensure stable operation and meet the aesthetic requirements for use in public places.


Stratos Explore Ultraleap demonstrator for robotics

The aim of this thesis is to systematically analyze the strengths and limitations of the Stratos Explore Ultraleap device in the context of controlling a robot. Subsequently implement a demonstrator application for showing off its applicability for robot control.


Continuous teleoperation setup for controlling mobile robot on streets

The task in this device is analyse available options for building a teleopearation cockpit for continuously controlling a mobile robot moving on the streets. The contribution of the thesis is to set up the system, validate its usability, and benchmark its capabilities/limitations on the ADL vehicle


ROBOTONT: ROS2 support for robotont

Creating ROS2 support for robotont mobile platform


Mixed-reality scene creation for vehicle teleoperation

Fusing different sensory feeds for creating high-usability teleoperation scene.


Validation study for AR-based robot user-interfaces

Designing and carrying out a user study to validate the functionality and usability of an human-robot interface.


Replication of the MIT Hydra demo

The goal of the thesis is to use the Hydra software package and integrate it for a robot used at IMS Robotics (e.g., TIAGo, Jackal, Robotont).
Hydra takes sensor data in input (stereo or RGB-D camera images and IMU data) and produces a hierarchical model of the environment, by estimating the trajectory of the robot (including loop closures), building a metric-semantic 3D mesh, and segmenting objects, places, and rooms in an indoor environment. These representations are combined into a 3D scene graph, which enables novel approaches for hierarchical loop closure detection and ensures the representations remain consistent after loop closure. Hydra is implemented in C++ and is ROS-compatible. It uses a neural-network based image segmentation front-end, but it otherwise relies on an efficient and multi-threaded CPU-based implementation, which is suitable for mobile robot deployment.
LINKS:
Video: https://youtu.be/qZg2lSeTuvM
Code: https://github.com/MIT-SPARK/Hydra
Paper: http://www.roboticsproceedings.org/rss18/p050.pdf


Quantitative Evaluation of the Situation Awareness during the Teleoperation of an Urban Vehicle

The goal of this thesis is to develop the methodology for measuring the operator's situation awareness while operating an autonomous urban vehicle. Potential metrics could include driving accuracy, speed, and latency. This thesis will be conducted as part of the activities at the Autonomous Driving Lab.


Completed projects

PhD theses

Masters's theses

Bachelor's theses