Text Topic B

Suggest Your Own Topic: Can’t find a suitable topic on the list? We always welcome your own topic ideas related to our research areas (e.g., human-AI interaction, socio-technical systems, digital transformation)! Simply contact Prof. Dr. Thomas Ludwig or the relevant team member.

 

Advisor: Prof. Dr. Thomas Ludwig

In VR training scenarios, virtual characters should not merely execute pre-scripted sequences, but should be able to react dynamically to situations and participants. Large language models can generate dialogues, action intentions, or structured states for an NPC. However, linguistic output alone is not sufficient for a credible simulation: statements and decisions must be accompanied by appropriate body postures, gestures, eye movements, and animations.

The goal of this work is to develop and investigate an architecture that can translate structured LLM outputs into appropriate avatar animations. In particular, the study will examine what information an LLM must provide, how this information can be reliably interpreted, and how animations can be synchronized with speech and behavior.

Broad research question:
How can the output of a large language model be automatically converted into believable and context-dependent avatar animations?

Key Areas

• Unity / Unreal Engine 
• Large Language Models 
• Avatar Animation 
• AI 
• NPC Systems

Advisor: Alexander Thier

Complex training simulations require a control center from which an operator can monitor and influence the ongoing scenario. With up to 16 participants taking part simultaneously, this gives rise to numerous possible functions: starting and ending scenarios, monitoring participants, triggering events, changing environmental conditions, controlling virtual characters, performing technical diagnostics, or flagging relevant situations for later debriefing.

This project will first examine what information and control options an operator actually needs and how these can be structured effectively. Building on this, a prototype UI and user interface concept will be developed and, if appropriate, integrated directly into the existing simulation platform. In addition to the implementation itself, the focus is particularly on information architecture, usability, and the avoidance of unnecessary complexity.

Broad research question:
What information and control options does an operator need to carry out complex multi-user VR training scenarios, and how can these be presented in a suitable interface?

Key Areas

• UI/UX 
• Operator Interface 
• Unity / Web UI 
• Multi-User VR 
• Requirements Engineering

Advisor: Alexander Thier

Manually creating new training environments is time-consuming and limits the number of different scenarios. A procedural content generation system could instead automatically or semi-automatically generate different operational environments from reusable building blocks. Possible examples include residential buildings, office spaces, warehouses, road sections, or other operational locations whose floor plans, rooms, objects, and hazard sources are generated based on configurable rules.

The initial goal of this work is to develop the basic architecture of such a system. The focus is less on the complete generation of highly complex operational scenarios and more on the question of how environments can be described modularly and generated reproducibly according to defined rules. Building on this, parameters such as room structure, doors, furnishings, operational objects, or hazard areas can be varied.

Broad research question:
How can a modular procedural content generation system be designed that automatically generates diverse, plausible operational environments suitable for VR training?

Key Areas

• Procedural Content Generation 
• Unity / Unreal Engine 
• Level Generation 
• Modular Assets 
• Rule-Based Systems

Advisor: Alexander Thier

Can't find a suitable topic on the list? We always welcome your own topic ideas related to our research areas (e.g., human-AI interaction, socio-technical systems, digital transformation)! Simply contact Prof. Dr. Thomas Ludwig or the relevant team member.

 

Advisor: Prof. Dr. Thomas Ludwig