abstract and a mental model needs to be created in order
to properly parse the information. Due to this, people are
often better at performing an action properly than teachinganother to do so. Therefore, if we find methods by which
to take teacher performance and apply it directly to students
motor learning systems, we may avoid the pitfalls normally
associated with teaching skills to another.
The goal of this system is to become a low latency, full-
time, highly parallel robotic motor skills teacher that can
provide constant motor-system feedback to the student as he
or she attempts to learn new motor skills.
B. Project Scope and Overview
The feedback system consists of four main modules
shown in Fig. 1. The first module shows the teacher and
student. The teacher performs a movement that the student
tries to mimic. Their performance is tracked optically by
a Vicon motion capture system. Results from tracking are
fed into our software that compares the performance of
the student and teacher to generate feedback commands for
the student. These feedback signals are then sent to the
wearable vibrotactile feedback suit, worn by the student.
Joints moving in error will receive vibrations proportional to
the amount of error. When the student’s body is in the right
position feedback is zero. Direct tactile feedback indicates
the discrepancy, wherever the student’s body is different
than the teacher’s. The rationale and implementation of this
system are presented next.II. MOTOR LEARNING
A. Motor Learning, Feedback, and Touch
Feedback is crucial to perform motor skills well [2], [3].
Performance is improved by both the specificity of feedback,
and its immediacy [4]. No other independent variable is
thought to affect one’s performance as much as immediate
feedback.
The skin is sensitive to many qualities of touch [5]. Skin
is specifically responsive to frequencies of roughly 250 Hz.
Vibrotactile actuator size influences frequency sensitivity.
Other factors such as low frequency oscillation (LFO) en-
velopes, sequences, and our adaptation to touch impulses,
are discussed at length in [5].We utilize an effect known as the ‘cutaneous rabbit’, more
formally known as sensory saltation [6]. A sequence of
properly spaced and timed tactile pulses will be processed as
if distributed “with more or less uniform spacing, from the
region of the first contactor to that of the [last].” This allows
us to utilize vibrotactile actuators as a means to directly com-
municate errors of joint angles. In addition, superimposing
saltative sequences can communicate rotational errors.
B. Previous work: Virtual Reality Training and other Tactile
Inventions
Virtual reality (VR) environments have been shown to
improve motor learning by providing augmented feedback.
Often VR displays overlay the student’s performance with
the desired movement, that provides an “intuitive and inter-
pretable form... sharing the same spatial frame of reference
[7].”
VR environments are useful because they emphasize the
differences between the subject and reference movements
and highlight desired trajectories in an understandable frame
[9]. In some complex tasks, VR training actually exceeds
training from a human expert [8]. VR training at times shows
more robustness to cognitive interference phenomena, such
as performing a task while being distracted with unrelated
behaviors [10]. Furthermore, VR training may be retained
longer than regular training [11].
Tactile actuators were originally developed for sensory
substitution, rather than for sensory augmentation. Prior
works have used tactile actuators to transmit speech infor- 动觉触觉交互学习系统英文文献和中文翻译(2):http://www.751com.cn/fanyi/lunwen_15697.html