Authors:
Summary:
Tactile feedback presents the most direct form of information. Learning a motor skill requires real time feedback. Unlike the auditory/ visual feedback, tactile feedback directly engages our motor learning system. The goal of this system is to become a low latency, fulltime, 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. Sensory saltation is a sequence of spaced and timed pulses. This helps to communicate the joint errors and rotational errors.
Vicon Tracking system used for subject tracking is a set of IR reflectors positioned at fixed points on the suit and a set of high speed IR camera to capture the position and joint angles. The markers’ placement and the calculated joint angles are used to find the five observed joints: wrist flexion/extension, wrist abduction/adduction, forearm rotation, elbow flexion/extension, and upper arm rotation. Calibration procedure is provided to change the joint angles and offsets. Control software compares the expert action with novice to give the tactile feedback. In the current system the error is determined by the joint angles. This needs to be improved to accurately gauge the performance of the user.
User study included 40 subjects divided into 2 groups. First 20 subjects were given only visual feedback and the next 20 subject were given both visual and tactile feedback. User is allowed to get used to the system by the first step. This step involves showing images of right hand positions to the users and asking the user to imitate the position as fast as possible. The vibrotactile feedback is given to the user for the first time as the user moves away from the position shown on the picture. The time taken for the user to attain the position is measured. The second step involves showing videos of actions (simple 1DOF - complex 5DOF). After 20 minutes of the video phase, the users are given a questionaire.
Results:
The tactile group felt the feedback didnot significantly help in improving performance. They made more conscious effort than the other group. The feedback for joint angle was better perceived than that for joint rotation. This is because the rotation feedback is given through saltation. This can also be seen in the results of the repeated trials and evaluating the amount learned.
Discussion:
Auditory and Visual feedback are the natural methods to get feedback. We have learnt to process the audio/ visual feedback while learning. Tactile feedback on the other hand is unnatural. Processing the feedback is a learning process in itself. This is also supported by the user study. The user group have had difficulties understanding the meaning of the tactile feedback.
The future work notes some interesting points - A better metric for comparison of the expert and novice has to be developed. Users should be allowed to correct one error dimension and then move on to the other (discomfort due to continuous tactile feedback).
In general the tactile modality is not a natural one for communication. However, in the case of motor movements with complex reaching actions it can be quite natural. As one reaches through an environment containing obstacles making mid course corrections based on feeling collisions by touch is quite common. Your point about correcting one error at a time is crucial. Multiple simultaneous stimulation would confound understanding and reduce precision.
ReplyDeleteI agree with the previous comment - I think vibrio-based feedback as a form of communication is going to be creepy for users. As a form of alerting (e.g. vibrating cell phones) or collision avoidance, users would likely be more comfortable with it.
ReplyDelete