Choosing a method of detecting and tracking the angle of each finger/joint is an important one. As this sensor will be wholly responsible for keeping the accuracy of the position for each joint. Any error can be amplified due to the serial nature of the hand.

The current sensors being considered are:

• Analog Hall effect angle sensor: ADA4571
• Flex Sensor: Flex Sensor 4.5 inch
• Potentiometer:
  • Rotary Position Sensor
  • Position Sensors/EVWAE/D

Additionally, the motors currently being planned for use as the actuators have integrated encoders, this will offer positional information on the motors.

Many robotic hands like this design use RC vehicle style servo motors which have a limited ark of movement. While they do vastly simplify the control system, as the control signal provides a desired angle and the servo then automatically try to reach that angle. No feedback is offered from RC servos, and so the current angle of the servo cannot be read without external sensors.

Using encoder motors the current position of the motor shaft can be accurately measured and there is no restriction to the angle of the shaft, including full rotations. This does however require the implementation of a dedicated control loop.

However, tracking the exact position/angle of the motor has debatable value, as neither closed loop servos nor encoder motors will offer the accuracy for measuring the current angle of each finger joint. Therefore, the encoder values may not be used as part of a closed loop control.

To detect the exact angle of each joint the above sensors are considered.

Hall effect sensors potentially offer the highest accuracy readings. However, interfacing with these sensors is by far the most complex and would require not only specific mounting for the sensor but also the integration of a magnet into each joint. The intolerance each sensor might have on nearby sensors is unknown, and potentially an issue.

The flex sensor would be by far the simplest to implement. A large resistor that changes resistance based off the angle, this sensor would be mounted along the spine of each finger and offer an approximate angle. This does mean tracking of individual joints would be lacking, and potentially precision would be as well. Also, the sensor would be laid along the back of each finger, very different from any other method suggested.

Potentiometers are simple resistor dividers, where the output voltage is varied based off the angle of an arm. In this use case the arm would be attached to the joint of each finger and so the angle could be measured using an analog to digital converter, most likely built into the microcontroller. This is currently the preferred method, as it is simple to implement, both in hardware and software, reliable and cheap. However, as a mechanical method it would be susceptible to wear after extended use, although this is predicted to be outside the scope of this use case. Another potential weakness would be the analog nature of the sensor’s feedback, increased length, or kinks in the wire could reduce tolerance, as well as the sensor signal being more susceptible to interference.