Robotics & Robot Perception
Robotics ROS Perception Human-Robot Interaction
Robots in the real world
My research has always circled back to physical systems. Before Quality-Diversity optimization and interactive AI became my core focus, I worked directly on robot perception, domestic robotics, and human-robot teamwork — building systems that had to function in messy, human-centred environments rather than clean simulations.
That background shapes how I approach optimization today: methods ultimately have to produce robust, adaptive behaviour at the interface of simulation, optimization, and real-world controllers.
b-it-bots — RoboCup@Home
As an active member of the b-it-bots team (2012–2015), I developed robotic perception systems for domestic robotics using ROS, C++, and Python. The team competed in the RoboCup@Home league, which challenges robots to assist humans in realistic home environments — fetching objects, recognizing people, and navigating cluttered spaces.
- RoboCup@Home selections with the b-it-bots team — perception systems for domestic robotics (2012–2015)
- RoboCup German Open, Magdeburg (2013)
- Team description paper: The b-it-bots Robo-Cup at Home 2014 Team Description Paper
Transparent object recognition
Within an independent R&D project, I developed a perception system capable of identifying transparent household objects based on the physical limitations of an RGB-D camera. Depth sensors typically fail on glass and other transparent surfaces; the approach fused multiple sensor modalities to recover reliable recognition.
This work was published at the RoboCup International Symposium 2016 and received the symposium’s Best Paper Award.
NIFTi — search-and-rescue robotics
At Fraunhofer IAIS (2011–2012) I contributed to NIFTi, an EU project on search-and-rescue robotics and human-robot interaction in urban search-and-rescue scenarios. NIFTi examined multi-robot deployments that enable cooperative teamwork and reduce the operator’s cognitive load — operationalizing natural collaboration as a trade-off between operational and cooperative demands within a cognitive architecture, minimizing cognitive load and optimizing joint team workflow.
Partners included the DFKI Language Technology Lab, TNO, Fraunhofer IAIS, Bluebotics, ETH Zurich, Czech Technical University in Prague, Sapienza University of Rome, the Dortmund Fire Department, and Corpo Nazionale Vigili del Fuoco (Italian Ministry of the Interior).
From robot controllers to optimization
Quality-Diversity optimization — the methodological core of much of my later work — originated in the robust control of robots. My interest in generating robust, adaptive controllers across simulation and physical systems is a direct continuation of this experimental robotics background, and connects to current work on multiphysical digital twins at the interface of simulation and physical robotic systems.