Guerrero, Roberto (2015-05). Low-cost Rigid-frame Exoskeleton Glove with Finger-joint Flexion Tracking Mapped onto a Robotic Hand. Master's Thesis. Thesis uri icon

abstract

  • This thesis provides a representation of a low-cost rigid-frame exoskeleton glove that is used to track finger-joint flexion mapped onto a robotic hand to mimic user movements. The overall setup consists of an exoskeleton glove (exo-glove), sensors, a microcontroller, and a telerobotic hand. The design of the exo-glove is crafted to fit onto a left hand. SolidWorks was used for the prototype designs which were then sent to the Stratasys 400 rapid prototyping machine to be 3D printed in ABS-M30 plastic. The exo-glove houses five rotary position sensors and three flexible sensors to track angle changes of the finger joints from two fingers and a thumb. Five low-pass filters are implemented as signal filtering for the rotary position sensors. An Arduino Mega microcontroller is connected to the sensors of the exo-glove and processes the input values. Using an open-loop controller to control the robotic hand, the values processed by the microcontroller from the exo-glove are sent to the servo motors on the robotic hand to operate the corresponding fingers of the user. Throughout the initial calibration and testing phase, each sensor was tested individually to ensure the sensor functionally performs well. Signal analysis was conducted on the sensors at steady state and while in operation to show fluctuations in sensor readings and response to finger flexion. Experimental results show that averaging sensor data in the processing code yields smoother values and better precision. Due to the use of low-pass filtering with the rotary position sensors, the data sets collected were grouped together tightly compared to the flex sensors without filtering. However, the actual angles measured were not accurately portrayed in sensor readings. The true flexion angles were compared in the data samplings to find a variety of ranges spanning around the angles desired to track. Many of the actual flexion angles were offset from the sensor readings by a variation of degrees, but the data shows the sensor readings were able to follow the general magnitude of the true flexion angles. The precision seen in the data was also apparent in the robotic hand mirroring the posture. Changes in sensor readings caused jerking movements to occur in the robotic fingers but were able to maintain an overall flexion mirroring of the RF exo-glove. There is quarter-second delay between the exo-glove sensor reading and the robotic hand mirroring capability when not implementing averaging. When averaging the sensor values, there was a delay of more than half a second between the exo-glove posture and robotic hand mirroring.
  • This thesis provides a representation of a low-cost rigid-frame exoskeleton glove that is used to track finger-joint flexion mapped onto a robotic hand to mimic user movements. The overall setup consists of an exoskeleton glove (exo-glove), sensors, a microcontroller, and a telerobotic hand. The design of the exo-glove is crafted to fit onto a left hand. SolidWorks was used for the prototype designs which were then sent to the Stratasys 400 rapid prototyping machine to be 3D printed in ABS-M30 plastic.

    The exo-glove houses five rotary position sensors and three flexible sensors to track angle changes of the finger joints from two fingers and a thumb. Five low-pass filters are implemented as signal filtering for the rotary position sensors. An Arduino Mega microcontroller is connected to the sensors of the exo-glove and processes the input values. Using an open-loop controller to control the robotic hand, the values processed by the microcontroller from the exo-glove are sent to the servo motors on the robotic hand to operate the corresponding fingers of the user.

    Throughout the initial calibration and testing phase, each sensor was tested individually to ensure the sensor functionally performs well. Signal analysis was conducted on the sensors at steady state and while in operation to show fluctuations in sensor readings and response to finger flexion. Experimental results show that averaging sensor data in the processing code yields smoother values and better precision. Due to the use of low-pass filtering with the rotary position sensors, the data sets collected were grouped together tightly compared to the flex sensors without filtering. However, the actual angles measured were not accurately portrayed in sensor readings. The true flexion angles were compared in the data samplings to find a variety of ranges spanning around the angles desired to track. Many of the actual flexion angles were offset from the sensor readings by a variation of degrees, but the data shows the sensor readings were able to follow the general magnitude of the true flexion angles.

    The precision seen in the data was also apparent in the robotic hand mirroring the posture. Changes in sensor readings caused jerking movements to occur in the robotic fingers but were able to maintain an overall flexion mirroring of the RF exo-glove. There is quarter-second delay between the exo-glove sensor reading and the robotic hand mirroring capability when not implementing averaging. When averaging the sensor values, there was a delay of more than half a second between the exo-glove posture and robotic hand mirroring.

publication date

  • May 2015