Until recently, the BRL carried out extensive yet largely independent research efforts in HRI. Building upon this existing expertise, the laboratory is in the process of structuring it into a strong cross-team HRI theme, aiming at establishing Bristol national and international leadership in Human-Robot Interaction. To reach this goal, the BRL recently recruited of a new professor who is to define and build up this novel cross-laboratory HRI theme.
Human-Robot Interaction covers a large academic spectrum, and requires an interplay between technical developments in robotics and several neighbouring academic fields like artificial intelligence and social psychology. As such, the BRL needs to broaden its interdisciplinary expertise to:
- structure and interface existing research efforts at BRL
- foster new interdisciplinary collaborations and projects with external actors
- open new funding opportunities
- gain visibility and recognition across the different academic fields related to HRI
As such, this document makes the case for opening a new senior position, complementary to the recently created professorship, and sketches the expertise and research vision that could be pushed forward.
Expected contributions
Federate the HRI efforts within BRL
As of today, 3 BRL groups carry out HRI-related research:
- the Safe human robot interaction, led by Alex Lenz
- Control for HRI, led by Angelika Peer
- Assisted Living, led by Praminda Caleb-Solly
Those groups produce excellent research, and should certainly not be merged in any way. Instead, it is about federating them around a common HRI theme.
Such federating effort can be effectively supported by a range of actions that I propose to take on:
- Defining a cross-BRL HRI 'brand' (including social media presence, logo and website) to which the different HRI-related groups can identify
- joint student supervisions
- write/coordinate joint funding proposals between groups
- creation of a master in HRI to complement UWE/UB master in robotics
- create and organise high-profile seminars
- define and setup a common technical platform for HRI activities, both hardware and software (ROS expertise)
Visibility within the HRI community
Beyond the BRL itself, I propose to support the international visibility and recognition of the BRL as one of the top academic research centre in HRI. I play since 2015 an important and recognised role within the Human-Robot Interaction community (Program Committee of the HRI conference in 2016 & 2017; theme chair in 2017; 10 publications at HRI over the last 4 years, including a best LBR and a Best Paper award).
I intend to leverage this position to bring visibility to the BRL, with the specific target to host the high-profile HRI conference in 2021.
Public visibility & outreach
I have an extensive experience in public engagement: nearly 20 years of science communication with major non-profit organisations like the French Planète Sciences, or the Paris and London Science Museums; media relations, including radios (Swiss national broadcast), TV broadcasts (BBC Arabic, Reuters, Swiss national TV), magazine/newspaper journalists; organisation and participation to major national/international events (eurobot, RoboCon, the South African SciFest festival, the European Robotics Week).
I plan to leverage this expertise to further develop public engagement in robotics with the public at local, national and international levels.
Attracting major new fundings for HRI at BRL
Network
- Current BRL partnerships:
- TUM (via Angelika Peer; Myorobotics)
- Fraunhofer IPA
- INRIA
- INTRO Marie Curie ITN: Umeå University, Sweden; Ben-Gurion University of
the Negev, Israel; Humboldt-Universität zu Berlin, Germany
Personal acquaintance with key research leaders in Europe: Tony Belpaeme, Angelo Cangelosi, Ana Paiva, Raja Chatila, Alessandro Saffiotti, Rachid Alami, Michael Beetz, Mohamed Chetouani, Herman Bruyninckx, several of the EPFL researchers (Auke Ijspeert, Aude Billard) and direct, personal contact with Francesco Mondada and Pierre Dillenbourg.
Supervisors Alami and Beetz are both leaders within the European robotics community, which directly benefited my engagement in the research community by allowing me to take part to selected symposia (like the 2010 Dagsthul seminar on HRI), to be introduced to other research leaders and to get involved into several large European projects (in particular, the FP7-CHRIS project with BRL)
New collaborations in the fields of assistive living (Auke Ijspeert, Alessandro Saffioti are active in this field), cognitive manufacturing (Rachid Alami, Michael Beetz have significant contributions in this field)
Experience in writing grant applications
- Successful application to the Marie Slodovska-Curie European fellowship (individual grant amounting for nearly 200 000 euros),
- taken active part to the writing of several other FP7/H2020 European projects, in particular the 'Rosely' project (modular robotics for elderly care), or the 'Scottie' project submitted this year
- also experience with national funding agencies, with proposals sent to the Swiss funding agency (National Swiss Fund)
Field Deployment Expertise
- over the last three years, I have conducted close to ten fields experiments with children and robots, providing me with key insights both on the methodological issues that arise in real-world child-robot interaction and on the interaction and acceptance factors of robots
- Within 3 years, major field experiment involving long-term deployment (> 3 months) of robots in a real-world elderly care centre, in close collaboration with Angelika Peer.
Technical Expertise
- extensive practical knowledge of hardware (robotic platforms: PR2, Nao, Pepper and several other custom platforms)
- software (important contributions to the ROS ecosystem, including core components -- support for Py3 -- and support for novel platform -- ROS for Nao), contributions to OpenCV, major open-source contributions -- MORSE, ORO, dialogs, gazr...
- and software engineering: many lectures and tutorials on ROS, software development, GIT, ontologies
- Unify the technical platform for HRI at BRL: one standard, ROS-based software stack, proper software engineering methods (software packaging)
- Quickly buy 2-3 Pepper robots (15 000 euros each) to be amongst the front-runner on this new platform which is likely to become the new golden standard for HRI (inexpensive, human-sized, designed with real-world deployment in mind, good mobility) -- in particular for assisted living applications.
Leadership
Leadership at EPFL
Previous experience: EPFL -> within two years, went from one single PhD student
working on psychological aspects of HRI within a lab of 30+ to a major research
group of the lab, with:
- 3 PhDs,
- 2-3 master students coming for the selective EU ViBOT master,
- 3-4 robotic platforms,
- joint supervisions with Portugal, France,
- 5 papers accepted at HRI, including a Best Paper Award and one Best LBR award,
- organization of workshops,
- major media coverage (BBC, Reuters, several radio + TV channels, > 30 local and international newspapers) for the CoWriter project,
- several journal papers including Robotics and Automation Magazine
Team leadership with MORSE
Team leadership skills in the context of MORSE:
- initiated the open-source project in 2008,
- lead developer
- rapidly took over the role of 'community manager',
- organizing several dedicated international workshops,
- designing and conducting many tutorials and hackathons,
- applying to (and receiving) external fundings for specific developments
As a result, MORSE went from a local, small-scale simulator project into a large open-source tool with hundreds of users (the MORSE publications have together more than 130
citations) and 40 contributors from 15 labs across the world.
Research vision
The research vision I would like to push forward at BRL spans over three directions. First, the development of a functionally complete yet practical 'Bristol Architecture for Social Interaction' with a focus on long-term robot autonomy. This would lead within two years to a long-term (> 2 months), high-visibility field experiment. Then, I would create a basic research track on cognitive architectures. I propose to explore a new hybrid symbolic/sub-symbolic paradigm build on top of associative memory networks. This would aim at developing a general and principled model of social cognition for robots. Finally, I would initiate a forward-looking, exploratory track on deep learning applied to human-robot interaction: applications of deep-learning to robotics are starting to emerge, and offer a promising framework to model the subtle dynamics of natural human-robot interaction.
These three dimensions feed each others, and build together an ambitious research agenda in cognitive robotics. I do believe that the Bristol Robotics Laboratory would be an ideal academic environment to carry it through.
1 - Complete and Practical Architecture for sHRI
- complete in the sense of covering all of the main surface functions of social interaction
- special focus on long-term autonomy
- practical in the sense of being focused on actionable principles, continuous deployment on the robots, with measurable outcomes: within 2 years, one major long-term field experiment in an elderly care centre, with expected high public visibility.
- strong commitment to open-source, standard-based development, engagement with the robotic community to share code and move forward quickly.
2 - Cognitive Architectures for Social Human Robot Interaction
The core of my current research agenda:
A few social behaviours that I would like a robot to exhibit:
- behavioural alignment -> surface alignment and/or global alignment -> the original 'maze' task by Pickering and Garrod is certainly relevant here: every time you play the same game with the same person, the interaction is more fluid
- ability to pass false-belief tasks, including non-physical, abstract ones,
- recursive awareness: being aware of being aware -- typically evidenced by being able to describe/verbalize its own state of awareness
- 'natural' turn-taking: I 'just' know when it is my turn to act
- 'natural' protodeclarative pointing: I 'just' know when I really need to draw your attention on something
The word 'just' is clearly the key here: these capabilities should be natural in the sense that they naturally follow from the model.
3 - Introduction of deep learning in HRI
- domains which require longer-term application, such as healthcare and education -> behave in an appropriate manner over longer timescales -> behaviours should be aligned with the humans' expectations.
- Generating such sustained and engaging social behaviours is an on-going research challenge. The recent progress of deep machine learning opens new perspectives that we should embrace to establish a leadership
- some momentum: RSS WS this year, special issue of IJRR 'Limits and Potentials of Deep Learning in Robotics', but HRI applications yet to be explored
- PinSoRo paper
Qualification
I hold a joint German-French PhD in cognitive robotics, focused on symbolic
knowledge manipulation applied to human-robot social interaction. I received for
this work the 'Best PhD in robotics 2012' award from the French National Centre
for Research (CNRS) as well as the highest possible mark ('Summa Cum Laude')
from the highly-regarded Technical University of Munich.
The PhD looked into characterizing and developing a semantic-aware cognitive
architecture for service robots, with a strong focus on cognitive interaction
with humans: ontologies, interactive symbol grounding, grounded dialogue,
perspective-awareness and Theory of Mind. My PhD work had as well a strong
experimental component, and I have systematically conducted studies in real
interaction scenarios with deployments on a diversity of platforms,
including complex robots like the PR2 or the BRL's 'Bert' platform.
Previously, I was awarded a joint German-French MSc in engineering (ENSAM
ParisTech/Karlsruhe Institute of Technology, completed in top 10 out of 1000
students). My final project looked at cognitive techniques for intelligent
manufacturing.
I also hold a MSc in Artificial Intelligence applied to education from Paris 5
University, completed with high honours.
Expertise
Since my PhD, I have significantly broaden my expertise on symbolic knowledge
for interaction both in the direction of artificial cognition for robots (with
publications in leading journals like 'Artificial Intelligence', publications
and awards in major international conferences including the RoMAN 2010 Best
Paper award, the AAAI 2015 Best Video award, and the HRI 2016 Best Paper award)
and experimental human-robot interaction, including several weeks-long
deployments of robots in schools and therapists' offices -- this was carried in
the frame of the EPFL CoWriter project which attracted in 2015 major media
coverage (40+ media outlets, including major ones like Reuters) and recently
resulted in a publication in the 'Robotics and Automation Magazine'.
My expertise in the field of HRI is recognized by my peers: I serve since 2015
as one of the 25 program committee members for the selective and
high-impact 'Human-Robot Interaction' (HRI) conference, and I have been invited
to serve as one of the theme chairs for HRI2017. I have also been awarded in 2015 the prestigious
Marie Slodovska-Curie European Fellowship for a project focused on the
challenges of social cognition for robots (actually entitled 'Donating Robots a
Theory of Mind').
My expertise is also recognized beyond HRI: besides my numerous technical
publications in top robotic conferences (ICRA, IROS, RSS), I act as Associate
Editor for the IROS conference; I am also regularly reviewing for the top
robotics journals (IJRR, IJSR, JHRI) and conferences (IROS, ICRA, HRI, RoMAN,
ICSR); I have been as well interviewed several times by international media
outlets (The BBC, the Fast Company, the Swiss National Radio, etc.) on questions
related to robotics and society.
According to Google Scholar, my 50+ publications have led to 860+ citations as
of today, with an H-index of 14.
Besides academic recognition, I also have a recognised technical expertise on a
range of robotic technologies. Over the last 10 years, I have directly worked on
a range of advanced robotic platforms, including autonomous cars, the PR2 robot,
the iCub, Aldebaran's Nao and Pepper robots, and several other custom robotic
platforms, like Munich's dual arms mobile manipulator 'Rosie'. I am a
contributor to the ROS core libraries and I have built up ROS support for
several new platforms, including Nao. I have contribution to the OpenCV computer
vision library, and I am the creator of MORSE, one of the leading robotic
simulator.
Team leadership
During my stay at the Computer-Human Interaction for Learning and Instruction
Lab (CHILI Lab, EPFL, led by Pierre Dillenbourg), I was effectively the
coordinator of the robotic group: I was tasked with growing the at-the-time
limited research activity in Human-Robot Interaction into a strong theme of the
lab. I have built a team that has now a recognized presence in HRI (amongst
others, a Best Late Breaking Report award at HRI2014; 7 full-paper accepted at
HRI over the last 4 years including a Best Paper award in 2016; and as
previously mentioned, several journal publications, including major venues like
'Robotics and Automation Magazine' as well as a lot of media coverage and public
attention for the lab). I was supervising at time up to 5 (MSc and PhD)
students, while playing in parallel an active role in the shaping of the
lab-level funding (writing of 2 grant applications, including European grants)
and research strategies.
Another testimony of my team leadership skills is my role in the 'MORSE
Simulator' open-source project: MORSE is a versatile robotic simulator that I
started at CNRS/LAAS in 2008. While I have been very much involved with the
development of the software from the beginning, I also rapidly took over the
role of 'community manager', organizing several dedicated international
workshops, designing and conducting many tutorials and hackathons, and applying
to (and receiving) external fundings for specific developments. As a result,
MORSE went from a local, small-scale simulator project into a large open-source
tool with hundreds of users (the MORSE publications have together more than 150
citations) and 40 contributors from 15 labs across the world.
Capacity to supervise and teach
To date, I have co-supervised 6 PhD students (4 are still on-going),
co-authoring so far 14 publications with them. As I currently co-supervise two
students in the UK, I am familiar with British regulations for PhD supervision.
I have also supervised numerous MSc students, almost systematically leading to
academic publications in major international venues.
In parallel, I teach since 2008, initially as teaching assistant, and more
recently as lecturer. I have conducted a variety of practical works in the
fields of computer sciences (teaching Ada, Java, SQL at undergrade and
postgraduate levels), artificial intelligence (Prolog, ontologies at MSc level)
and mechatronics. More recently, I have co-designed at EPFL a new undergraduate
module on visual computing. I proposed and led ambitious practicals involving 3D
rendering, computer vision and augmented reality. In 2016, I have been
lecturing the 'Humanoids and Mobile Robots' undergrade module at Plymouth
University, which covers a range of subjects, including Kalman filtering,
bipedal walking, localisation and mapping and robot control architectures. I
taught as well several guest lectures at postgraduate level at Toulouse
University, EPFL and Plymouth University on subjects including ROS, ontologies,
RGB-D vision for HRI or the principles of software engineering. Finally, I have
been invited as faculty to the Social HRI Summer School, where I was teaching on
visual attention tracking in HRI.
Capacity to establish collaborations and attract external fundings
Through my involvement in the different dissemination activities previously
mentioned (like the HRI Programme Committee) and several European projects (most
notably, the CHRIS FP7 project, which also involved the BRL), I have built a
dense network of colleagues in the fields of cognitive robotics and human-robot
interaction. This network is particularly developed in Europe, and besides
excellent relationships with Prof. Rachid Alami, Prof. Michael Beetz, Prof.
Pierre Dillenbourg and Prof. Tony Belpaeme with whom I worked directly, I enjoy
close relationships with some of the key European research leaders in HRI and
service robotics like Prof. Ana Paiva, Prof. Alessandro Saffioti, Prof. Raja
Chatila, Prof. Auke Ijspeert, Prof. Aude Billard, or Prof. Mohamed Chetouani. I
do believe I can build upon this network to establish successful new
collaborations in the fields of assistive living (Auke Ijspeert, Alessandro
Saffioti are active in this field), cognitive manufacturing (Rachid Alami,
Michael Beetz have significant contributions in this field), and of course,
human-robot interaction.
I do also have the experience of grant application writing, in particular
European grants: besides my successful application to the Marie-Slodovska Curie
European fellowship (individual grant amounting for more than 195 000 euros), I
have taken active part to the writing of several other FP7/H2020 European
projects, in particular the 'Rosely' project (modular robotics for elderly
care), tentatively submitted by EPFL in 2014, or the 'Scottie' project,
submitted this year. As such, I have practical experience with the process of
international proposals writing and submission.