Article

Prosthetic components and control interfaces for upper limb amputees
have barely changed in the past 40 years. Many transradial prostheses have been
developed in the past, nonetheless most of them would be inappropriate if/when a
large bandwidth human-machine interface for control and perception would be
available, due to either their limited (or inexistent) sensorization or limited
dexterity. SmartHand tackles this issue as is meant to be clinically experimented
in amputees employing different neuro-interfaces, in order to investigate their
effectiveness. This paper presents the design and on bench evaluation of the
SmartHand. METHODS: SmartHand design was bio-inspired in terms of its physical
appearance, kinematics, sensorization, and its multilevel control system.
Underactuated fingers and differential mechanisms were designed and exploited in
order to fit all mechatronic components in the size and weight of a natural human
hand. Its sensory system was designed with the aim of delivering significant
afferent information to the user through adequate interfaces. RESULTS: SmartHand
is a five fingered self-contained robotic hand, with 16 degrees of freedom,
actuated by 4 motors. It integrates a bio-inspired sensory system composed of 40
proprioceptive and exteroceptive sensors and a customized embedded controller
both employed for implementing automatic grasp control and for potentially
delivering sensory feedback to the amputee. It is able to perform everyday
grasps, count and independently point the index. The weight (530 g) and speed
(closing time: 1.5 seconds) are comparable to actual commercial prostheses. It is
able to lift a 10 kg suitcase; slippage tests showed that within particular
friction and geometric conditions the hand is able to stably grasp up to 3.6 kg
cylindrical objects. CONCLUSIONS: Due to its unique embedded features and
human-size, the SmartHand holds the promise to be experimentally fitted on
transradial amputees and employed as a bi-directional instrument for
investigating -during realistic experiments- different interfaces, control and
feedback strategies in neuro-engineering studies.

Langue : ANGLAIS