Interdisciplinary student teams are supervised in cross-departmental collaboration to develop AI solutions for real-life use cases. Each team is supervised by an AI pilot and an application pilot. The pilots first implement the application-specific HW and SW, which enables the students to develop their own solutions. Here, the balance between the challenge and the solution elements provided on the AI and application side is balanced in such a way that it is appropriate for the respective student target group. This ensures the necessary flexibility for the curricular anchoring in the participating disciplines.
In the "Radar epilepsy detection" application area, an AI is being developed that uses radar data to attempt to detect an epileptic seizure. A labeled data set from Munich University Hospital already exists that can be used for this purpose. As part of a project, the data is to be processed and an AI designed and developed.
In the "Reinforcement Learning Contest" application area, teams compete against each other in a virtual environment, each with their own reinforcement solution for their artificial agents based on a known or their own reinforcement learning algorithm. The performance of the individual agents will be automatically evaluated in a virtual contest arena. All fully executable student projects are awarded a certain number of credits. The winning teams in each semester are honored in a (virtual) Hall of Fame.
In the "Industrial robotics" field of application, an AI application and an associated physical demonstrator are being developed and set up in the field of industrial robotics based on a real application from the series production of a medium-sized company. The swivel movements performed by the robot are learned by an AI in a simulation environment, checked by another system and forwarded to a robot that performs the movements. In the project, students work in interdisciplinary teams to gain insights into reinforcement learning techniques as well as kinematics and robot path planning.
Further link: Laboratory for Digital Product Development and Manufacturing: RoboKI
In the "Electrophysiology" application field, an AI is being developed that converts eye movements into mouse movements and a prolonged blink into mouse clicks. The eye movements are recorded and processed using electro-oculography. The AI is intended to achieve greater precision in the recognition of blinking and eye movements. Data collection, planning and development of the AI are to be implemented as part of a project.
In the "Mobile robot" application field, an AI is being developed for a robot that can recognize previously defined objects and push certain bodies into an area. The robot has several sensors that help with spatial orientation. Camera, laser, sonar and lidar data can be accessed. Well-known algorithms, such as SLAM (Simultaneous Localization and Mapping), already exist for implementation and can be used.
In the "FaSiMo" field of application, the dynamic driving simulator is to be developed for AI applications. As part of their project work, students will design, develop and test an AI-supported function. They can choose from two subject areas, which will give them an insight into the application of AI systems in the automotive sector.
In traffic sign recognition, students develop a driver assistance system that uses AI methods to recognize traffic signs in a scene. The system can then react accordingly, e.g. by adapting the speed to new specifications.
The Takeover Time Prediction project uses a variety of data sources, such as eye tracking and interaction behavior, to estimate the time and quality of takeover requests during automated driving.
Pulse-coded neurons (SNN) are to be used in the "Spiking Neural Networks" application area. In contrast to classic AI, the neuron models used are very similar to biological neurons in terms of signal generation and signal transmission. These models can be simulated by software on the one hand, and there is also hardware designed for this purpose. The aim of the field of application is to carry out experiments with the SNN and compare them with the classic methods of AI (machine learning). Typical application scenarios are image recognition and speech recognition.
In the "traffic object detection" application area, a system is being further developed and used to generate data for traffic planning. In traffic planning, quantitative information about road users (pedestrians, cyclists, various types of vehicles, etc.) is recorded in spatial and temporal aggregation. One approach is to extract this information from digital video images using object recognition. In the project, students gain an insight into practice-oriented image processing using AI systems by collaboratively developing and applying a system for traffic object detection. They always work in interdisciplinary teams of civil engineers and computer scientists.
Marvin Schneider, M. Sc.
Marv.Schneider(at)hochschule-trier.de
Schneidershof | Gebäude O | Raum 5
Eileen Neumann, B. Sc.
eileen.neumann(at)hochschule-trier.de
Schneidershof | Gebäude O | Raum 5
You are leaving the official website of Trier University of Applied Sciences
