Systems, Man & Cybernetics - January 2016 - 16

many cases, their capabilities are
MPH based on the work de still limited.
scribed in [14] but focused more
Compared to the
on the advances of intermanual
◆ They only work with SPHs, and
original intention of
interactions. Further details of
their interaction models supthe system were discussed in [5].
port only SPH. This is a fundahaving multimanual,
In [22], [23], the authors proposed
mental limitation.
multifinger haptic
string-based multimanual, mul◆ They focus only on kinesthetic
tifinger haptic systems and
haptics and leave nowhere in
devices that support
implemented a number of protheir frameworks to support
both kinesthetic and
totypes. In [24], [25], the authors
tactile haptics.
tactile feedbacks,
implemented a series of biman◆ They enumerate off-the-shelf
ual MPHs with five fingers on
devices with known I/O DoF
single-point kineseach hand.
and provide limited support
thetic haptic devices
4) Haptic includes both kinesthetic
for customized or augmented
and tactile feedback, yet in many
devices.
(SPHs) are a comprocases, they are considered and
◆ There are no definitions of varimise between intuiimplemented separately. There
ous interaction models. Augmenttive interactions and
have been a number of efforts to
ing existing dev ices with
promote holistic haptics. In [26],
intermanual, interfinger, or interaffordability.
the author analyzed the tradisensory interactions has not been
tional way of kinesthetic haptics
systematically studied and in
and proposed a conceptual
many cases can be very difficult.
implementation for kinesthetic and tactile haptics. In
There are also a few efforts in standardizing the interac[27], the authors proposed to modify an existing SPH
tion models for haptic systems. In [32], a joint initiative
into a device combining kinesthetic and tactile feedwas proposed to define an International Organization for
back. In [28], a design that couples kinesthetic and tacStandardization (ISO) standard for future haptics, includtile feedback together for shape and texture was brought
ing kinesthetic and tactile. Although there has been no
forth. In [29], the authors presented a systematic design
detailed implementation of the ISO standard yet, the idea
of integrating kinesthetic and tactile feedback in a modsounds very promising.
ular approach.
Although much work has been done and many aspects
Principles of Multipoint Haptics
of MPHs have been covered, most previous works focus on
By analyzing various existing and future haptic device
the hardware design and implementation. The software
requirements and understanding the limitations of the
framework design perspective has not drawn equal attenexisting haptic frameworks, we have identified a number
tion. Until now, it is still unclear how different MPHs and
of design principles for an abstract MPH model that covers
their underlying interaction models should uniformly
all possible configurations of MPHs.
communicate and collaborate to promote MPH applica1) The number of manuals and the number of fingers (which
tions to a wider range and deeper involvement of realcan be considered equivalent to HIPs) on an MPH should
life scenarios.
be flexible rather than just fixed for rigid configurations.
There have been a number of noticeable attempts to
Although currently we only have bimanual MPHs and on
promote device-independent platforms for SPHs, such as
each manual the number of HIPs only ranges from one to
HAPTIK [30] and CHAI3D [31]. HAPTIK is an SDK that sits
five, we should not restrict the possibility of more manubetween hardware and high-level haptic libraries. It is built
als and more HIPs on each manual for the accomplishon top of the component object model (COM) and uses
ment of more dextrous and complex tasks.
plug-ins to provide low-level access to off-the-shelf SPHs. It
2) The input and output DoF on HIPs should be defined in a
is also supposed to work with existing and future kinesconsistent way rather than just being enumerated. Curthetic SPHs. CHAI3D is another widely used haptic SDK
rently, the most commonly adopted input DoF for haptic
because it is open source and supports devices from differdevices are either three (translation only) or six (translaent vendors. In many cases, CHAI3D and its standard
tion and rotation), with the addition of one or two butdemos are used for benchmarking haptic collision and
tons for the device rather than for any individual HIP.
rendering algorithms. It integrates visual rendering and
This assumption could not fulfill the requirement of
collision detection algorithms for quick implementations
existing and future MPHs, as each HIP may involve an
and also includes a physics engine and a deformation
independent array of vibrotactile sensors as well as sepaengine for more complicated simulations.
rate on/off switches in addition to translation and rotaAlthough HAPTIK, CHAI3D, and other similar previous
tion input. Similarly, the commonly adopted output DoF
works are popular within the haptics community, in
for current haptic devices are either three (force only) or
16

IEEE SyStEmS, man, & CybErnEtICS magazInE Janu ar y 2016

Table of Contents for the Digital Edition of Systems, Man & Cybernetics - January 2016