Modulcode:	infMPInS-01a
Englische Bezeichnung:	Master project - Intelligent Systems
Modulverantwortliche(r):	Prof. Dr.-Ing. Sven Tomforde
Turnus:	unregelmäßig (WS20/21 SS21 WS21/22 SS22 WS22/23 SS23 WS23/24 SS24)
Präsenzzeiten:	4PÜ
ECTS:	10
Workload:	300 h project work, from which 60 h are in presence
Dauer:	ein Semester
Modulkategorien:	MSc-Inf-Proj (MSc Inf (21)) 2F-MSc-Proj (2F-MA Inf (21)) Proj (MSc Inf (15))
Lehrsprache:	Englisch
Voraussetzungen:

Kurzfassung:

An "intelligent system" is a computing system capable of operating under difficult conditions (e.g. time-varying environments, emergent situations or disturbances) by autonomously adapting its behaviour to changing conditions and learning autonomously. We use the "Turtle Bot 3" robot platform as the basis for developing intelligent systems in student projects.

Lernziele:

The participants should learn to work on a larger, cohesive project from the field of intelligent systems in a limited time. The focus is on independent project management and group work.

The overall goal of the course is to derive a basic understanding and experiences in how intelligent systems are designed, developed, and operated. Besides the actual development of the software, the module aims at practical experiences in a team of developers. Particular goals are:

• Specification and definition of "product" using standard software engineering tools
• (Self-)Organisation as a team, management of the process, deadline supervision
• Entire software development process until delivery to the customer
• Documentation of the product

Lehrinhalte:

The actual teaching content changes according to the topic. Check the group website for this year's topic.

Summer term 2024: For the visual inspection of road signs, we have several sequences of images in which a sign is recognised and identified as the same sign across the sequences. This results in a selection of images of the same object, which will now be used to assess its condition. There are several potential approaches. One possibility is to develop a solution that automatically selects the best single image. This can then be analysed for damage or other deviations from the ideal state using an anomaly detection method. Another approach would be to fuse information from different images before anomaly detection. The project group will implement and evaluate strategies for data selection or fusion as well as compare methods for visual anomaly detection. The addition of your own ideas to the suggestions described here is explicitly encouraged. The goal of the project is therefore to develop a system that can analyse road signs for visual anomalies using a selection of images.

Winter term 2023/24: The goal of this project is to evolve a controller for a TurtleBot robot that is based on weight agnostic neural networks. This is a method to find neural network architectures with a strong inductive bias to solve a specific task. In contrast to the usual approach of optimising a fixed network's weights, here the architecture is optimised by an evolutionary algorithm while all connections share a single weight value. The TurtleBot has to navigate several race tracks, whose borders can be detected by the robot's LIDAR, without collisions. Different objectives can be targeted, such as short lap times, particularly well interpretable controllers, or a good performance on diverse test tracks. The project includes the setup of a simulation environment, the implementation and optimisation of weight agnostic neural networks in that simulation, the transfer to real robots, and the evaluation of the approach on self-built race tracks with real robots.

Summer term 2023: The aim of the project is to develop a robust system that enables a mobile "TurtleBot3 Burger" robot to classify the type of surface it is driving on, even at different speeds. By correctly classifying it into known classes, the robot can make an informed decision regarding an appropriate driving style (e.g. lower speed or avoid the area). Time series of measurements from the robot's proprioceptive sensor system, in this case an inertial measurement unit (IMU), serve as the basis for the classification.

Winter term 2023: The aim of the project of this term is to develop a condition monitoring system for two TurtleBot3 Burger robots working on a repetitive collaborative task. As a result, the team of students will present a binary-output programme, which immediately signalizes an alert in case of detecting any anomaly during the robots run. This programme should be able to distinguish between robots' inner failures and outer unexpected events. The last ones should have a zero or as low as possible anomaly scores.

You'll investigate the scenario as a team and develop the software in a customer-developer setting. The expected outcome consists of i) the software (ideally as a ROS module), ii) a (very brief) technical documentation, and iii) a report (as a scientific paper following the IEEE template in conference mode, sumamrising the research hypothesis, the approach, and the achieved results).

Fundamentals of the module Besides the actual development of the software, the module aims at practical experiences in the following aspects:

• Specification and definition of a 'product' using standard software engineering tools
• Setup of scientific experiments using standard Data-Science libraries like numpy/pandas/keras/tensorflow/...
• (Self-)Organisation as a team, management of the process, deadline supervision
• Entire software development process until delivery to the customer
• Documentation of the product

Your supervisors are continuously monitoring the process in the role of a customer, i.e. you are expected to regularly demonstrate the progress.

Weitere Voraussetzungen:

No mandatory requirements. However, prior knowledge in the field of "intelligent systems" is more than helpful, ideally achieved by attending the corresponding lectures.

You should have good programming skills and no fear of working with real hardware.

Prüfungsleistung:

Presentations, report, scientific paper, and the realised software system (incl. documentation).

Lehr- und Lernmethoden:

Project work (practise).

Verwendbarkeit:

Literatur:

For the individueal project at the beginning of the project.

Verweise:

The KickOff Meeting is scheduled for: Wednesday, April 17, 2024, at 14:15 in HRS 3, Room 309b (INS Lab). In case you can't attend this meeting please contact Nils Bischoff via email nib@informatik.uni-kiel.de) - participation is also possible without attending the KickOff meeting.

See current task description of the term at: http://www.ins.informatik.uni-kiel.de/en/teaching/master-project

Interested students may contact the group (i.e., Prof. Tomforde, st@informatik.uni-kiel.de) for more details. You can also do this to "book a place" in the project already during the preceeeding term.

Kommentar:

This project is offered as a master project (If you are interested in a Bachelor/Master Thesis within the scope of this project, please contact st@informatik.uni-kiel.de for more information)

There will be a KickOff Meeting at the begin of the lecturing period. See website of the group for details.