Invited Speakers

Luis MerinoUniversidad Pablo de Olavide (UPO), Spain

Title: Learning from demonstrations as a methodology for human-aware robot navigation
Abstract: Human comfort in the presence of robots is influenced by the social awareness capabilities of such robots when performing their tasks. In particular, when robots move among persons, aspects like social constraints, legibility, human affordances and others have to be taken into account. A methodology to incorporate these aspects into the navigation stack is learning from demonstrations. The talk will present different methods for learning social navigation behaviors from data, and will discuss how these methods can be used to transfer socially-appropriate behaviors to the robot from direct demonstrations or by observing how people move. The talk will also discuss how crowd simulation models can be also used to incorporate human-like navigation behaviors into robotic agents. 

Biography: Luis Merino is Associate Professor of Systems Engineering and Automation and Principal Investigator of the Service Robotics Laboratory of the Universidad Pablo de Olavide (UPO), Seville, Spain. His research interests include autonomous navigation, motion planning and control for social robotics, perception, and planning under uncertainties and multi-robot systems. He has published more than 70 papers, and has led UPO’s team in several national and international projects, on those topics. He serves as Associate Editor for the Image and Vision Computing Journal and conferences like ICRA and IROS, where he has also co-organized several workshops on the aforementioned topics. 

Francesco Ferro, CEO of PAL Robotics
TitleDesigning and using robots for a safe Human-Robot cooperation
AbstractPAL Robotics' mission is to develop service robots that enhance people's quality of life. To achieve this, it is key that robots can operate safely around (and in collaboration with) people. For this reason, one of our main priorities in ensuring that our robots are intrinsically safe at all levels: from its design to the components and the robots' behaviour. This talk will provide insights on how we develop safe robots and show applications in which PAL Robotics' platforms interact with people responsively in different scenarios.

BiographyFrancesco Ferro is the CEO and co-founder of PAL Robotics, one of the top service robotics companies around the world, and a euRobotics aisbl Board Director. He received a BSc+MSc degree in Telecommunications Engineering in 2002 at the Politecnico di Torino (Italy), a Master at ISEN in Lille (France) and an Executive MBA at the University of Barcelona (Spain) in 2011. Since 2004 he is developing cutting-edge humanoid service robots at PAL Robotics. This robotics company based in Barcelona has the mission of making people’s life easier by using robotics. 
For more than 14 years the team has developed robots for service tasks and Industry 4.0, as well as for R&D. TALOS is their latest creation, standing as one of the most advanced and powerful biped humanoids for industry demands. TALOS joined the PAL Robotics’ family in 2017, also composed by TIAGo, a collaborative robot for Industry 4.0 systems; the TIAGo Base for courier applications and logistics; and StockBot, the robot that automates inventory in retail stores; amongst other cutting-edge humanoid platforms such as REEM-C for R&D purposes and REEM as a service provider in dynamic environments.

Dikai Liu, University of Technology Sydney
TitleAn Intelligent Robotic Co-worker for Human-Robot Collaborative Blasting: Enabling Methodologies, System Design and Practical Applications  
Abstract: Physically intensive tasks and hazardous working environments are very common in many industry sectors. A typical example is abrasive blasting which requires workers to spend long periods of time resisting large forces and to adopt awkward body postures in very dusty and low visibility environments. Supplementing manual labour with robotic aids will have significant health, safety and economic benefits. There has been increasing interest throughout the world in the use of assistive robotic technology. However, a number of key research challenges need to be addressed before robotic systems can be deployed to physically assist human operators with varying physical sizes and strengths, and work in typical unstructured industrial environments. This presentation will discuss the research on how robots can be enabled to physically assist diverse workers performing labour intensive tasks such as abrasive blasting, and the development of an intelligent robotic co-worker: Assistance-as-Needed roBot (ANBOT). Topics include (1) the enabling methodologies developed in a 5-year research project, including a model-based adaptive scheme for estimating assistance need, human-pose recognition in dusty environments, physical human-robot interaction and safety framework; (2) system design of ANBOT: (3) experimental results from field trials; and (4) deployment strategies.

BiographyDr Dikai Liu is a distinguished professor and Director of the Centre for Autonomous Systems at the University of Technology Sydney (UTS), Australia. His main research interest is robotics with the focuses on infrastructure robotics, robot teams and physical human-robot collaboration. Besides conducting fundamental robotics research, he has led the development of autonomous and collaborative robots that can be practically deployed in real applications, including autonomous robots for steel bridge maintenance, bio-inspired climbing robots for inspection of complex steel structures, intelligent robotic co-worker for human-robot collaborative abrasive blasting, smart hoist for patient transfer, and Intervention AUVs (I-AUV) for underwater structure maintenance. Since 2005, his research has received 9 research and engineering excellence awards at international, national, New South Wales state and university levels. His research also received three Best Paper Awards from international conferences. He is the recipient of many patents include USA and Australia patents.

Anca Dragan, University of California, Berkeley

Title Safe prediction and learning in human-robot interaction

Abstract: In this talk, I will advocate that robots should treat human actions as information about what they want: this enables robots to better anticipate human motion during collaborative tasks, as well as to learn the learn how people prefer to be assisted. However, I will also highlight that this can fail spectacularly when people have unmodeled biases or care about something completely outside of the robot's hypothesis space. Imagine for instance a person deciding to avoid some spilled coffee on the floor that the robot has no idea about -- we don't want to the robot to stubbornly think that the person will turn and walk straight through the coffee any second, despite their behavior so far. In our recent work, we found that this "misspecification" can actually be detected online. Human actions will start seeming irrational to the robot, which will then adjust its model and make higher-variance predictions about the future, thereby becoming more conservative and maintaining safety. 

Biography: Anca Dragan is an Assistant Professor in EECS at UC Berkeley, where she runs the InterACT lab. Her goal is to enable robots to work with, around, and in support of people. Anca did her PhD in the Robotics Institute at Carnegie Mellon University on legible motion planning. At Berkeley, she helped found the Berkeley AI Research Lab, is a co-PI for the Center for Human-Compatible AI, and has been honored by the Sloan fellowship, the NSF CAREER award, the Okawa award, MIT's TR35, and an IJCAI Early Career Spotlight.