Current projects
CAROUSEL+ (2021 - 2024)
Dance with real people and AI-driven characters in an online Virtual Reality space. The goal is to combine novel interaction technologies with physics-based character animation, to make people able to feel each other’s presence, touch, and movement, even if they are not in the same physical space.
MOSIM (2018 - 2021)
The aim of MOSIM is an end-to-end digital integration based on modular simulation of natural human motions. Within the European economy, digital modelling activities and the simulation of human motion in particular, have emerged during the last decades in various domains ranging from automotive and truck over healthcare, construction and pedestrian simulation to the gaming industry. Even though differing in their respective scope, the ability to realistically predict real-world observations has shown to be a key technology in order to remain competitive. For mechanical and mechatronic components, this trend is already covered by various research projects related to smart factories. In contrast, the generation of a rich repertoire of realistic human motions in complex and possibly highly collision afflicted environments is not sufficiently addressed by commercial tools, yet. Moreover, complex process workflows with an exhaustive number of possible manual task sequences can only be partly addressed today, since process variants have to be modelled by hand. As manual modelling is inevitably linked with additional effort, the potential cost reduction is significant. In order to introduce approaches and software solutions, which are capable to automatically simulating a rich repertoire of realistic human motions, MOSIM aims to develop and implement a generic concept which is inspired by the Functional Mock-up Interface (FMI) standard. MOSIM transfers the idea of co-simulating models from different simulation environments to the field of human simulation by means of introducing the Motion Model Units (MMU).
REACT (2017 - 2020)
The goal of REACT is to use simulations to generate training data for the control of autonomous cars via deep neural networks. Our group works on the simulation of predestrians with large stylistic variations using deep learning based motion synthesis techniques.
Hybr-iT (2016 - 2019)
The project aims at developing and implementing worker robot collaboration scenarios in the manufacturing industries. Our group works on the visualization of the human agents for the simulation of the scenarios and investigates physics based motion synthesis for handling collisions.
INVERSIV (2014 - 2017)
In the context of Industrie 4.0, the project developed a system for validating the design of cyber-physical systems (CPS). As part of the project different designs of CPS were explored using modules for semantic sensor data analysis, hybrid automata, and agent technologies for representing human activities and interactions for maintaining production. Our group applied a statistical motion synthesis approach for worker simulation.
INTERACT (2013 - 2016)
The project INTERACT investigated the simulation of assembly work places in the manufacturing industry for ergonomic evaluations. Our group developed a statistical motion synthesis method that was integrated with a text-based user interface developed by DFKI Language Technology Lab and an ergonomic analysis solution by project partner imk automotive.