Wearable Robots: A Grand Challenge in Robotics

Speaker: Harry Asada , MIT MechE

Date: Monday, November 18, 2013

Time: 4:00 PM to 5:00 PM Note: all times are in the Eastern Time Zone

Public: Yes

Location: 3-270

Event Type:

Room Description:

Host: Domitilla Del Vecchio, MIT MechE

Contact: John J. Leonard, jleonard@csail.mit.edu

Relevant URL:

Speaker URL: None

Speaker Photo:
None

Reminders to:

Reminder Subject: TALK: Wearable Robots: A Grand Challenge in Robotics

Imagine that one day, you have a third arm and a
third leg attached to your body. The extra limbs
will help you hold objects, support your body,
share a workload, and streamline the execution of
a task. If the movements of such Supernumerary
Robotic Limbs (SRL) are tightly coordinated with
your own arms, you may come to perceive the extra
limbs as an extension of your body, incorporated
into your body image. The objective of this work
is to develop key enabling technologies for
transforming robots to act as parts of a human
body. Wearable SRLs are opening up new horizons
of robotics, posing diverse research issues and
challenges ranging from machine design and
human-robot coordination, to biomechanics, motor
control, and machine learning and perception.

Three types of Supernumerary Robotic Limbs being
built at the d'Arbeloff Lab will be presented: 1)
a lightweight robot sitting on the shoulder of a
human for lifting and supporting objects in the
overhead area, 2) a seven-fingered hand (5
fingers + 2 robotic fingers) for grasping and
manipulating large/odd-shaped objects, and 3) a
pair of wearable canes attached around the waist
for supporting and bracing the human body. For
these wearable robots, communication and
coordination with the human is the key
challenge. Three aspects of coordination control
will be presented. First, the concept of
biological synergies is applied to the
seven-fingered hand in order to control the two
robotic fingers in concert with the five human
fingers. Through grasp experiments and data
analysis using Principal Component Analysis it
will be shown that synergies exist for
seven-fingered hands as well as for five-fingered
hands. For real-time control, Partial Least
Squares (PLS) regression is used for extracting
control laws from the data that can best
correlate the posture of the two robotic fingers
to that of the five human fingers. Second,
interactive human-robot-task processes are
modeled as a concurrent, distributed event system
based on Coloured Petri Net (CPN). A type of
hybrid control system is constructed by replacing
CPN's static state transitions with dynamic,
proactive transition laws learned from human
demonstration data. Finally, a learning algorithm
inspired by biological muscle training is applied
to the wearable canes. Untrained robotic
actuators are treated as un-innervated muscles.
Repeated exposures to simultaneous physical and
informational stimulations lead to formation of
artificial neuromuscular junctions that control
the wearable robots. The seminar will be
concluded with potentials of wearable SRLs and
their social and economic impacts, ranging from
increasing productivity and safety for factory,
construction, and field workers to improved
quality of life for elderly people and the
handicapped as well as reduced workload for
caregivers and clinicians.

Research Areas:

Impact Areas:

This event is not part of a series.

Created by John J. Leonard at Tuesday, November 12, 2013 at 6:48 PM.