BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Hierarchical Multi-Agent Reinforcement Learning through Communicat
 ive Actions for Human-Robot Collaboration - Elena Corina Grigore (Yale Uni
 versity)
DTSTART:20161214T150000Z
DTEND:20161214T160000Z
UID:TALK69619@talks.cam.ac.uk
CONTACT:Louise Segar
DESCRIPTION:As we expect robots to start moving from working in isolated i
 ndustry settings into human populated environments\, our need to develop s
 uitable learning algorithms for the latter increases. Human-robot collabor
 ation is a particular area that has tremendous gains from endowing a robot
  with such learning capabilities\, focusing on robots that can work side-b
 y-side with a human and provide supportive behaviours throughout a task ex
 ecuted by the human worker. In this paper\, we propose a framework based o
 n hierarchical multi-agent reinforcement learning that considers the human
  as an “expert” agent in the system—an agent whose actions we cannot
  control but whose actions\, jointly with the robot’s actions\, impact t
 he state of the task. Our framework aims to provide the learner (the robot
 ) with a way of learning how to provide supportive behaviours to the exper
 t agent (the person) during a complex task. The robot employs communicativ
 e actions to interactively learn from the expert agent at key points durin
 g the task. We use a hierarchical approach in order to integrate the commu
 nicative actions in the multi-agent reinforcement learning framework and a
 llow for simultaneously learning the quality of performing different suppo
 rtive behaviours for particular combinations of task states and expert age
 nt actions. In this talk\, we present our proposed framework\, detail the 
 motion capture system data collection we performed in order to learn about
  the task states and characterise the expert agent’s actions\, and discu
 ss how we can apply the framework to our human-robot collaboration scenari
 o.
LOCATION:CBL Room BE-438\, Department of Engineering
END:VEVENT
END:VCALENDAR