Bachelor and Master theses

In the following list are topics for both Master or Bachelor theses, please get in touch with the contact person listed, to find out more (last update April 2019).


Topic: High-bandwidth self-organizing network architecture for distributed robot skin. Design of a distributed interface nodes for robot skin.

Contact person: Florian Bergner

-------------

Topic: Constrained visual servoing. Fusion of multi-modal tactile skin with visual servoing control.

Contact person: Emmanuel Dean

-------------

Topic: Visual tracking with high-speed eyes.

Contact person: Emmanuel Dean

-------------

Topic: Force-drift compensation using temperature sensors for the multi-modal robotic skin.

Contact person: Florian Bergner

-------------

Topic: TOMM: sensor fusion using correlated information from Multi-modal exteroceptive and interoceptive information.

Contact person: Florian Bergner

-------------

Topic: TOMM picking challenge. Integrate results obtained by the RoboCup@Home team in the TOMM robot.

Contact person: Emmanuel Dean

-------------

Topic: Full multi-modal sensor calibration using proprioceptive information (forces, proximity, joint and visual information )

Contact person: Florian Bergner

-------------

Topic: Stride cancellation in presence of unexpected obstacles for robust and stable bipedal walking. (Stable emergency stop during walking)

Contact person: Rogelio Guadarrama

-------------

Topic: Adaptive End-effector dynamic compensation: obtaining dynamic model of manipulated objects.

Contact person: Emmanuel Dean

-------------

Topic: Embedded Robot skin driver. Implement a driver for the robot skin for embedded systems.

Contact person: Florian Bergner

-------------

Topic: Distributed energypacks (analysis of energy flow to determine the requirements of batteries and power grids in robotic systems)

Contact person: Florian Bergner

-------------

Topic: Robot modelling toolbox for C++

Contact person: Emmanuel Dean

-----------------

Topic: Robots that are able to distinguish their own body: sensorimotor learning algorithm for robotic self-distinction Self/other distinction, sensorimotor contingency learning, sensorimotor self, multisensory perception, enactive robots

Contact person: Dr-Ing. Pablo Lanillos

Description: The robot should be able to distinguish its own body from other agents and inanimate objects in order to interact with the environment. For this purpose, the robot learns the relation between the sensors and the actuators (i.e., spatio-temporal sensorimotor correlations). This should involve more than one sensor modality such as touch, vision and proprioception. This is a quite novel approach within the embodied artificial intelligence and gets inspiration from robotics, artificial intelligence and cognitive sciences. The study will be mainly performed on the humanoid robot TOMM developed at ICS. It consists of two UR-5 arms and a base (compact omni-directional mobile platform), running under ROS. It is almost covered with artificial skin and has stereo and RGBD vision systems.

The thesis is associated with the SELFCEPTION Marie Skłodowska-Curie project. For more info please visit: www.selfception.eu. Tasks and Objectives The Master thesis project goal will be adapted depending on the background and motivations of the student (control, machine learning, etc)

General goals:

1. Implement a system in ROS able to grab multisensory information from the robot using the available interfaces (from the artificial skin, the joint positions and the visual data).

2. Design an algorithm that learns the relation between sensors and actuators (regressor, connectionist, Bayesian inference, ...).

3. Design an on-line algorithm that learns the sensorimotor contingencies during robot interaction.

4. Implement self-exploration and interaction algorithms.

Prerequisites:

Programming skills in C++ and Robotic Operating System (ROS)

Probabilistic, connectionist or regressors modelling knowledge is a plus.