Gesture Controlled SMA Actuated Robotic Hand
Embedded Files
This project aims to present an innovative interface between a Shape Memory Alloy (SMA)-based robotic hand and a glove equipped with flex sensors. SMAs are chosen as actuators as they have an excellent force-to-weight ratio and produce no noise. The Nickel-Titanium SMA has proven to be one of the most beneficial materials in actuation applications owing to its high power-to-weight ratio and high achievable actuation strains. The project’s human-machine interface (HMI) component utilizes a glove equipped with flex sensors. These sensors capture user gestures, and a machine learning algorithm interprets this data, enhancing gesture recognition accuracy. The classified gestures are then wirelessly transmitted to the robotic hand, forming a seamless link between user intent and robotic action.
The hand was designed such that the palm houses all the Ni-Ti springs while also providing space for the organization of wires.
A Support Vector Machine Model (SVM) machine learning model was used to classify the gestures shown by the user, with an accuracy of 99.7 %. This interpreted gesture was wirelessly translated to the robotic hand.
The figure above depicts an experiment in which the robotic hand was employed remotely to activate a wire cutter.
The robotic hand was designed to be capable of bi-actuation, employing two sets of five springs.
Outcomes
Patent published and under examination - Application IN202421051320
Page updated
Google Sites
Report abuse