guy hoffman

co-director, media innovation lab
interdisciplinary center herzliya

p.o.box 167
herzliya, israel
hoffman ϑ idc°ac°il
My research is concerned with physical action and joint action in humans and robots. I try to develop computational models of physical practice ("motor learning") and teamwork and implement them on robotic and other artificial agents. This serves both to inform psychological accounts of practice, and to design agents that are more efficient, and more adaptive when the repeatedly perform a task. read more

In particular, I study anticipation and its role in perception and collaboration, fluency between human and robotic team members, perceptual-symbol based cognition for robots. Other projects include hybrid-control puppeteering for robotic stage and film actors, and human-guided machine learning.

list of publications

selected research projects

Huamn-Robot Joint Musical Performance

How can humans and robots perform live music together, in a fluent and synchronized way, even if the music is structured and precise? In this work I extend my anticipatory approach to time-critical human-robot musical performance. Shimon, a robotic marimba player listens to a human pianist, and plays it part in a a jazz duo.
Video: Annotated highlights
Video: Full performance

Anticipatory Perceptual Simulation

Anticipatory Perceptual Simulation is a novel approach to robotic agent cognition, allowing a robot to adaptively train on a task in conjunction with a human teammate. During joint practice with a human, the robot increasingly anticipates the human's activity and uses it in form of perceptual simulation or "hallucination". We have implemented this framework on two robots, and - in human-subject studies - find the anticipatory behavior not only more efficient, balanced, and fluent, but that subjects find robots with this behavior more intelligent and contributing compared to a reactive robot.
Paper: AAAI'08
Video: Anticipatory Perception

Anticipatory Action in MDPs

Anticipation is a key ingredient in solitary action, and more importantly in joint action and collaboration. We have developed a generic anticipatory action framework for Markov Decision Processes, implemented it on a computer game agent, and show how anticipation improves efficiency, fluency, and trust when tested with untrained subjects.
Papers: T-RO'07 HRI'07 (Best Student Paper)

Robotic Desk Lamp

AUR is a robotic desk lamp, a collaborative lighting assistant. It serves as a non-anthropomorphic robotic platform for studies of human-robot fluency, embodiment, and nonverbal behavior. AUR won the 2007 IEEE International Robot Design Competition.
Link: More Info on AUR
Video: AUR

Human-Robot Theater

We are developing the groundwork for robotic theater actors. These are built on principles from acting theory, and pave the way for fully autonomous robotic actors. We have developed a hybrid control system as a first stage, and staged a first-of-a-kind performance including two human and one robot actor.
Papers: RO-MAN'08 AI-50
Video: Robot Theater

Human-Guided Reinforcement Learning

Sophie's Kitchen is an online research platform in which untrained humans guide an agent that learns through reinforcement learning. We studied how people want to teach agents, and how reinforcement learning needs to change to fit human's expectation and tendencies.
Paper: RO-MAN'06, HRI'06 (Best Student Poster)

Collaboration in Human-Robot Teams

A task planning and execution framework for an autonomous humanoid robot working shoulder-to-shoulder with a human teammate, sharing the workspace and the objects required to complete a task. We implemented the use of shared goals, dynamic meshing of sub-plans, and non-verbal behavior.
Paper: AIAA'04 (Best Paper in Session)

The Case for Non-humanoid Robots

From a design perspective, there is a good reason to explore non-humanoid forms for personal robots. By steering away from imitating the human form, the design space opens up infinitely. By embellishing familiar objects — such as household items, tools, appliances, or furniture — we can encourage people to reconsider their relationships with the artifacts surrounding them.
Thesis Chapter: RO-MAN'08

this is the year 2010, but there's nothing copyrighted here.

news:


talk:
"designing human-robot performances" @
MIT Media Lab, Cambridge, MA, USA
01 jun 2010

press:
shimon featured on The Colbert Report
12 may 2010

award:
ICRA 2010 Best Paper Award for Cognitive Robotics
6 may 2010

talk + paper:
"Gesture Based Human-Robot Jazz Improvisation" @
International Conference on Robotics and Automation (ICRA'10)
3-8 may 2010, anchorage, ak, usa

press:
shimon featured on CNN.com
29 apr 2010

press:
business week blog calls shimon's DLD-2010 opening performance "highlight of the evening"
25 jan 2010

performance:
shimon's international preimiere at DLD 2010 in munich. gala concert and live demo all through the conference.
24-26 jan 2010
video excerpt, photos, press to follow

press:
shimon's jazz improvisation and new head on npr's all things considered.also here and here.
22 dec 2009

demo:
shimon at siggraph asia. also: gizmodo and previous post
14 dec 2009

writing:
guest post on "so where's my robot?"
1 dec 2009

award:
ro-man 2008 cognitive human-robot interaction best paper award
16 oct 2009

paper:
"Effects of Anticipatory Perceptual Simulation on Practiced Human-Robot Tasks"
accepted to 'Autonomous Robots' Journal, Springer press
28 sep 2009

award:
awarded a virginia center for the creative arts residency
02 aug 2009

talk:
"practicing with robotic teammates and stage partners" @
intel labs, seattle, wa
07 aug 2009

performance:
"p.i.d gestures" human-robot jazz improv
17 apr 2009 @ eyedrum, atlanta, ga
watch on youtube

design:
illustraion commissioned by
the new york times "week in review"
15 feb 2009

panel:
"israel goes hollywood - media without borders" @
the israel conference
4 jun 2009, los angeles, ca

new video:
"citycast" overview
20 jan 2009 @ idc herzliyah
watch on youtube

talk:
"practice, anticipation, and embodiment for robots and agents" @
technische universität münchen
20 feb 2009, munich, germany

talk:
"practice, anticipation, and embodiment for robots and agents" @
georgia tech gvu (TSRB 132)
12 feb 2009, atlanta, ga

talk:
"designing personal robots" @
bezalel israeli academy of art and design
31 dec 2008, jerusalem, israel

talk:
"building robots for everyday interaction" @
interdisciplinary center herzliyah
30 dec 2008, herzliyah, israel

talk:
"judging helpless machines: how people assess robots" @
univ. of texas austin (SEA 1.332)
8 oct 2008, austin, tx

talk:
"human-robot interaction" @
11th annual chinese-american kavli frontiers of science
2-4 oct 2008, beckman center, irvine, ca

talk:
"designing personal robots" @
nokia research center
4 sep 2008, palo alto, ca

talk + paper:
"A Hybrid Control System for Puppeteering a Live Robotic Stage Actor" @
17th International Symposium on Robot and Human Interactive Communication (RoMan'08)
1-3 aug 2008, munich, germany

talk + paper:
"Anticipatory Perceptual Simulation for Human-Robot Joint Practice: Theory and Application Study" @
23rd AAAI Conference on Artificial Intelligence (AAAI'08)
13-17 jul 2008, chicago, il