-
ViT-VS: On the Applicability of Pretrained Vision Transformer Features for Generalizable Visual Servoing
Authors:
Alessandro Scherl,
Stefan Thalhammer,
Bernhard Neuberger,
Wilfried Wöber,
José Gracía-Rodríguez
Abstract:
Visual servoing enables robots to precisely position their end-effector relative to a target object. While classical methods rely on hand-crafted features and thus are universally applicable without task-specific training, they often struggle with occlusions and environmental variations, whereas learning-based approaches improve robustness but typically require extensive training. We present a vis…
▽ More
Visual servoing enables robots to precisely position their end-effector relative to a target object. While classical methods rely on hand-crafted features and thus are universally applicable without task-specific training, they often struggle with occlusions and environmental variations, whereas learning-based approaches improve robustness but typically require extensive training. We present a visual servoing approach that leverages pretrained vision transformers for semantic feature extraction, combining the advantages of both paradigms while also being able to generalize beyond the provided sample. Our approach achieves full convergence in unperturbed scenarios and surpasses classical image-based visual servoing by up to 31.2\% relative improvement in perturbed scenarios. Even the convergence rates of learning-based methods are matched despite requiring no task- or object-specific training. Real-world evaluations confirm robust performance in end-effector positioning, industrial box manipulation, and grasping of unseen objects using only a reference from the same category. Our code and simulation environment are available at: https://alessandroscherl.github.io/ViT-VS/
△ Less
Submitted 6 March, 2025;
originally announced March 2025.
-
UGV-CBRN: An Unmanned Ground Vehicle for Chemical, Biological, Radiological, and Nuclear Disaster Response
Authors:
Simon Schwaiger,
Lucas Muster,
Georg Novotny,
Michael Schebek,
Wilfried Wöber,
Stefan Thalhammer,
Christoph Böhm
Abstract:
Robotic search and rescue (SAR) supports response teams by accelerating disaster assessment and by keeping operators away from hazardous environments. In the event of a chemical, biological, radiological, and nuclear (CBRN) disaster, robots are deployed to identify and locate radiation sources. Human responders then assess the situation and neutralize the danger. The presented system takes a step…
▽ More
Robotic search and rescue (SAR) supports response teams by accelerating disaster assessment and by keeping operators away from hazardous environments. In the event of a chemical, biological, radiological, and nuclear (CBRN) disaster, robots are deployed to identify and locate radiation sources. Human responders then assess the situation and neutralize the danger. The presented system takes a step toward enhanced integration of robots into SAR teams. Integrating autonomous radiation mapping with semi-autonomous substance sampling and online analysis of the CBRN threat lets the human operator localize and assess the threat from a safe distance. Two LiDARs, an IMU, and a Geiger counter are used for mapping the surrounding area and localizing potential radiation sources. A mobile manipulator with six Degrees of Freedom manipulates valves and samples substances that are analyzed by an onboard Raman spectrometer. The human operator monitors the mission's progression from a remote location defining target locations and directing the semi-autonomous manipulation processes. Diverse recovery behaviours aid robot deployment, system state monitoring, as well as recovery of hard- and software. Field tests showcase the capabilities of the presented system during trials at the CBRN disaster response challenge European Robotics Hackathon (EnRicH). We provide recorded sensor data and implemented software through a GitHub repository: https://github.com/TW-Robotics/search-and-rescue-robot-2024.
△ Less
Submitted 20 September, 2024; v1 submitted 20 June, 2024;
originally announced June 2024.
-
Autonomous Vehicle Calibration via Linear Optimization
Authors:
Georg Novotny,
Yuzhou Liu,
Wilfried Wöber,
Cristina Olaverri-Monreal
Abstract:
In navigation activities, kinematic parameters of a mobile vehicle play a significant role. Odometry is most commonly used for dead reckoning. However, the unrestricted accumulation of errors is a disadvantage using this method. As a result, it is necessary to calibrate odometry parameters to minimize the error accumulation. This paper presents a pipeline based on sequential least square programmi…
▽ More
In navigation activities, kinematic parameters of a mobile vehicle play a significant role. Odometry is most commonly used for dead reckoning. However, the unrestricted accumulation of errors is a disadvantage using this method. As a result, it is necessary to calibrate odometry parameters to minimize the error accumulation. This paper presents a pipeline based on sequential least square programming to minimize the relative position displacement of an arbitrary landmark in consecutive time steps of a kinematic vehicle model by calibrating the parameters of applied model. Results showed that the developed pipeline produced accurate results with small datasets.
△ Less
Submitted 27 April, 2022;
originally announced April 2022.
