Workshop 3

Interdisciplinary Advances and Challenges of AI, Robotics and Vision (InterAC 2024)

Modern robotics is an interdisciplinary field integrating concepts from Vision and Artificial Intelligence (AI). Robotics is also a major driving force for fundamental research on Vision and AI as well as for the development of novel applications in these interrelated areas.

This workshop aims to bring together researchers from academia and industry to discuss the latest developments, challenges, and open research questions. Contributions to the workshop are solicited on the following topics::

Software Design & Architecture
Safety, Security and Standardization in Robotics
Estimation, Diagnosis and Learning
Field and Rescue Robots
Robot Sensing and Perception, Machine Vision
UAV, UGV, Autonomous Systems, Service Robots
Reasoning Methods for Robotics
Advanced Industrial Robotics
Motion Planning and Control
Collaborative Robotics, Human Robot Interaction
Modelling and Design in Robotics
Humanoids, Legged, and Bioinspired Robots
Autonomous Mobile Manipulation
Embedded AI in Robotics
Deep Learning for Embodied Systems
Deep Reinforcement Learning
Movement Primitives and Motor Skills
Data-Driven Methods for Robotics and Vision
Machine and Deep Learning for Embodied Systems

Program

Invited Speakers / Talks

Human Intention Reading Robots

Univ.-Prof. Dr. Dongheui Lee, TU Wien, Institute of Computer Technology Robots are becoming increasingly integrated into our lives, interacting with us and assisting us in various tasks. To ensure effective interaction and collaboration between humans and robots, it is essential that they understand our intentions and anticipate our actions. In this talk, I will introduce some of the approaches we developed. Our team developed an efficient and robust transformer-based model to detect and anticipate Human-Object Interactions (HOIs) from videos. This enhanced anticipation empowers robots to proactively assist humans, resulting in more efficient and intuitive collaborations. We showcase the effectiveness of our approach through benchmark using VidHOI dataset and experimental results in a real robot.

Visual and Tactile Robot Learning

Univ.-Prof. Dr. Elmar Rueckert University of Leoben, Cyber-Physical-Systems

Robot Learning focuses on the ability to learn complex skills through learning from demonstration, via self-improvement aka Reinforcement Learning, and most recently as well, via contrastive learning. Latest fundamental achievements in deep learning enabled researchers to utilize complex sensory data such as tactile receptor data, RGB video streams or depth camera measurements. In this talk, I will present our latest results in that direction with respect to tasks in robot grasping and manipulation, world model learning, representation learning and robot navigation in dynamic environments.

“The Human Touch” - Exploring Human Perception of Robots and Closing the Loop in Human-Robot Interaction

Ass.-Prof. Dr. Astrid Weiss, TU Wien, Informatik

Understanding and predicting people’s reactions to robots is a multifaceted challenge prevalent across diverse contexts, including service robotics, social robotics, and collaborative/industrial robots. In this talk I want to highlight the complexity of this issue, emphasizing the significance of use case development. The challenge lies in effectively translating user requirements into robot designs that align with human expectations and needs. Furthermore, replicating study results poses a hurdle, demanding a nuanced approach due to the variability in human responses. Beyond controlled environments, predicting real-world consequences adds another layer of complexity, requiring comprehensive considerations of societal, ethical, and practical implications. Addressing these challenges is pivotal for the successful integration of robots into various domains, but also for the research field of Human-Robot Interaction.

Contributed Talks (15 min + 5 min questions):

Encouraging Exercise in Dementia Patients: A Reproduction Study with Robot Pepper as Dance Therapist (ID: 3)
SCARAB - Autonomous Demand-Oriented Waste Disposal with Mobile Robot (ID: 15)
Workspace Registration and Collision Detection for Industrial Robotics Applications (ID: 32)
Improving 2D-3D Dense Correspondences with Diffusion Models for 6D Object Pose Estimation (ID: 34)
GP FastSLAM - Simultaneous Localization and Mapping using Gaussian Process Robot Kinematics (ID: 35)
Enhancing Transparent Object Pose Estimation: A Fusion of GDR-Net and Edge Detection (ID: 36)
Unsupervised Learning of Effective Actions in Robotics (ID: 38)
New Forms of Human-Robot-Collaboration (ID: 39)
Artificial trust-based task allocation method in multi-human-robot teams (ID: 40)

Poster Pitches (2 mins):

Reinterpretation of the Non-Slippage Impact Direction for Elastic Contact Transitions (ID:11)
Automated Electric Vehicle Charging System: A Robotic Approach for Seamless Plug Connection with Deep Neural Networks (ID: 12)
Towards Affordance-Based Explanations of Robot Motion Planning (ID: 13)
LiDAR-IMU calibration on a ground vehicle (ID: 18)
Intelligent Data Monitoring: A Combination of Rule-Based and Machine Learning Approach (ID: 26)
Towards an Accepted Service Robot for Retail 4.0 (ID: 30)
Adversarial Framework for Monitoring Unannotated Anomalies of Key Performance Indicators in Robot Applications (ID: 31)
A Combined Approach for Robot-based Disassembly and Testing of Electric Vehicle Battery Packs (ID: 33)
Chance & Curiosity: How Does Action Noise Exploration Compare to Curiosity-Based Intrinsic Rewards? (ID: 37)
Optimizing Performance of Robots with Closed Control Architecture: Deep Learning-Enhanced Model-Based Control (ID: 41)