Systems, Man & Cybernetics - January 2016 - 15

he two hands and ten fingers that humans
are born with are the most dexterous parts of
the body and the intrinsic tools that have created our history. To advance human-computer interaction so that we make use of such
dexterity, an intuitive interface is highly desirable. Work in
this field has led to the era of haptic interaction, yet the
advances of haptic hardware and the corresponding software development kits (SDKs) have remained at the singlepoint stage for many years. Compared to the original
intention of having multimanual, multifinger haptic devices that support both kinesthetic and tactile feedbacks, single-point kinesthetic haptic devices (SPHs) are a
compromise between intuitive interactions and affordability. They are limited for probing-like operations and incapable of performing more sophisticated tasks. This limitation
has caused the SPHs to lose much of their value in real-life
applications where the dexterity that human hands possess is essential. Another reason for the slow transition
from SPHs to multipoint haptic devices (MPHs) is the
unique hardware designs and software implementations
for MPHs. Unlike SPHs that share a rather similar architecture that can be easily abstracted for common communications, MPHs can be significantly different in their
looks and the underlying interaction models, such as a
multimanual model for collaborative hand manipulations
and a multifinger model for pinch and grasp. The situation
becomes even worse when considering incorporating both
kinesthetic and tactile feedbacks together for more comprehensive systems, as these two are usually considered
separate under existing hardware and SDK implementations, while they actually both belong to haptics and are
interdependent on human hands.
In this article, we propose a universal MPH framework
with its underlying abstract MPH model, which support
existing and future multimanual, multifinger haptic systems, rendering both kinesthetic and tactile feedbacks.
The abstract MPH model defines multiple interaction models hierarchically, which can be helpful in the definition of
hardware scope for customized haptic systems. The
framework defines the primitive components and operations, which can be adopted by various drivers and SDK
configurations for advanced MPH applications.
Related Works
Most SPHs have three or six input-output (I/O) degrees of
freedom (DoF). They are effective at performing tasks
such as probing and gliding but are incapable of more
sophisticated tasks that require multiple hands and fingers, such as advanced applications in the field of medicine [1], [2], industry [3], and design [4]. Previous works

have identified the necessity and superiority of MPHs over
SPHs. This includes enhanced operational experience to
existing applications and innovative operational modalities to more challenging applications [5]-[7].
Theoretically, MPHs should be multimanual and multifinger and should include both kinesthetic and tactile
feedback for each haptic interaction point (HIP). Yet such
implementations require extensive research across a number of research topics. Therefore, most existing implementations focus only on part of these components. We
categorize existing implementations below based on their
involved interaction models.
1) Multifinger systems focus on the representation of interfinger operations such as pinch and grasp. A typical
implementation approach for such systems is to combine
multiple SPHs as a single MPH, where the interfinger
communications are actually inter-SPH communications. This approach has been applied to multiple offthe-shelf haptic devices, such as the Sensable Phantom
Premium [8], Force Dimension [9], and Novint Falcon
[10], but these systems usually face the drawback of
reduced workspace. Another approach is to augment an
existing SPH with multifinger capabilities while keeping
its original kinematics. In [6], [11], [12], the authors have
extended the Sensable Omni and Force Dimension into
cable-driven multifinger devices and implemented a new
interaction model. There are also efforts in building new
multifinger devices from scratch, such as the work
described in [13]-[15].
2) Multimanual systems focus on the representation of intermanual operations and usually omit interfinger interactions with simplified hand models. The majority of
multimanual systems are bimanual systems while, theoretically, a higher dimension of manuals is also possible
and can be useful in many applications. A multiple SPH
combination approach is also popular, such as the work
discussed in [2], [16]. Typical device combinations are
homogeneous, although heterogeneous configurations are
also possible, such as the work discussed in [17], where a
Novint Falcon and a Phantom Omni were combined. There
are also other efforts in building new bimanual SPHs
from scratch, such as those described in [18], [19].
3) A number of research efforts have been made to create
more capable MPHs involving both multimanual and multifinger features. This has led to further advances in task
completion and human-machine interaction. In [20], the
authors have proposed a bimanual, multifinger grasping
system. It involves two augmented Phantom Premium
devices, each having four motors and seven sensors to
perform grasp and manipulation. In [21], the authors proposed MasterFinger, a customized multifinger, bimanual
Ja nu a r y 2016

IEEE SyStEmS, man, & CybErnEtICS magazInE

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