Program
The workshop will be held the 26 of november 2020, starting from 10:30.
10:30 Introduction
The workshop organizers
Session 1: Service and Industrial Robotics
Chair: Salvatore M. Anzalone
10:35 Jonathan Cacace, Riccardo Caccavale and Alberto Finzi
Supervised Hand-Guidance during Human Robot Collaborative Task Execution: a Case Study
10:45 Angelo Ferrando, Zeid Kootbally, Pavel Piliptchak, Rafael C. Cardoso, Craig Schlenoff and Michael Fisher
Runtime Verification of the ARIAC competition: Can a robot be Agile and Safe at the same time?
10:55 Discussion
Session 2: Social Robotics
Chair: Luca Boncompagni
11:05 Lucrezia Grassi, Carmine Recchiuto and Antonio Sgorbissa
Knowledge-driven conversation for social robots: exploring crowdsourcing mechanisms for improving the system capabilities
11:15 Mattia Barbaresi, Stefano Bernagozzi and Andrea Roli
Robot choreographies: artificial evolution between novelty and similarity
11:25 Gloria Beraldo, Luca Tonin, Amedeo Cesta and Emanuele Menegatti
Shared approaches to mentally drive telepresence robots
11:35 Discussion
Session 3: Assistive Robotics
Chair: Alberto Finzi
11:50 Jianling Zou, Soizic Gauthier, Thomas Gargot, Dominique Archambault, David Cohen and Salvatore Anzalone
Development of a Wizard-of-Oz based framework for QTrobot to facilitate the learning of handwriting in children with dysgraphia
12:00 Luca Buoncompagni, Alessio Capitanelli, Marta Cristofanini, Antonella Giuni, Fulvio Mastrogiovanni, Carola Motolese, Andrea Nisticò, Alessandro Sperindè and Renato Zaccaria
Persuasive and Polite Sentences to Drive Human-Robot Interaction in Smart Homes for Elderly Care
12:10 Discussiuon
12:20 Lunch
Session 4: Keynote
14:00 Lorenzo Jamone – Queen Mary University of London, UK
Toward intelligent robots: haptics and dexterity
The robots of today are mainly employed in heavy manufacturing industries (e.g. automotive): these are big robotic manipulators which perform simple and repetitive tasks in very structured environments. The robots of the future will be different. They will perform more complex tasks in more complex unstructured environments, even in collaboration with humans. They will be more intelligent machines. How will this be achieved? Explicit insights from biology and psychology, well established control and engineering principles, modern AI techniques, have to be combined and properly integrated. In the talk I will briefly summarize my main research interest in the area of Cognitive Robotics, with a focus on physical robotic intelligence: humanoid robots that learn representations of their own bodies and of the external environment through interactive exploration, and that can eventually display problem solving capabilities in unstructured settings. Then I will quickly discuss my current activities at QMUL, mostly centered on haptics and dexterity (the intelligence in the hands!): tactile sensing, grasping and manipulation, haptic feedback.
14:30 Questions
Session 5: Planning and Scheduling
Chair: Alberto Castellini
14:35 Luca Biferale, Michele Buzzicotti, Fabio Bonaccorso, Patricio Clark di Leoni and Kristian Gustavsson
Optimal control of point-to-point navigation in turbulent flows using Reinforcement Learning
14:45 Giulio Mazzi, Alberto Castellini and Alessandro Farinelli
Policy Interpretation for Partially Observable Monte-Carlo Planning: A Rule-Based Approach
14:55 Marin Lujak and Mislav Matezovic
On Efficiency in Dynamic Multi-Robot Task Allocation
15:05 Discussion
Session 6: Panel session
Chair: Alberto Castellini
15:20 Verification and anomaly detection for intelligent robots
Francesco Amigoni, Alessandro Farinelli, Luca Iocchi
Ensuring correct, secure and safe operations for intelligent robots is a key challenge to build trustworthy AI systems. This is particularly important when such systems can be deployed in open and dynamic environments and must interact with humans. A main element to improve robustness and, eventually, to build trust in these systems is the ability to monitor their operations by detecting anomalous behaviours and avoiding unsafe actions or wrong decisions. Anomaly detection techniques and verification approaches can be used to achieve these aims and are subject of a growing interest from industries and from the research community. Differently from other computer systems, robots can be equipped with decision making components and physical actuators that can be used to actively improve the assessment of the anomaly and implement proper defense actions. In this roundtable we will briefly introduce key topics related to anomaly detection and verification for intelligent robots from an AI point of view, arguing that a shift from passive to active anomaly detection and verification could be beneficial. We will then have an open discussion, focusing on which problems are more challenging and interesting, which techniques can be used to address these problems, in which scenarios such techniques could have a significant impact and above all how the AIRO community can contribute in this research area.
15:55 Conclusions
The workshop organizers