​Sensor Challenges 

We had difficulty in controlling the sensors and getting consistent, reliable data. Most of our time was spent on testing sensors and calibrating them to get the desired values. Having a slightly different positioning of the flex sensor and accelerometer would sometimes invalidate the data collected. We have to be extremely careful with the sensors. Flex sensors were sensitive to movement of the fingers or movement of the gloves.  When the they were sewn onto the gloves, certain parts buggle up (due to different finger lengths and the knuckle location). Over time, with constant testing and transporting them around, one of the flex sensors broke at the base. Luckily, we had a few spare flex sensors and was able to replace, re-solder and resew the broken flex sensor back on the glove. 

Even though the accelerometer was working, it was also sensitive and noisy that we had to tweak the code multiple times to ensure that it was working consistently.  To reduce the noise, we wrote code that would only return an average acceleration value over multiple interval points (20) and over a period of time (1 second). In order to get a consistent value for capturing a specific gesture, we decided to detect gesture by capturing a huge change in one axis. Initially, we had code to detect four different types of gestures: moving the accelerometer “left and right”,  “back and forth”, “in a circle”, and “up and down”. However, we ended up only using one gesture to reduce the engineering demands on another part of the project.

Issues with Buttons

We had a problem of detecting button clicks. The user sometimes has to click the buttons multiple times before getting a response. We believe that this is due to the delay in the loop section of our program. The loop section contains/calls many lines of code, and delays naturally as the program processes the code. This resulted in increasing the time interval before checking for clicks, which leaves clicks often go undetected. We didn’t use the button library for this project because we thought having additional button gestures (such as press hold) was unnecessary. However, using the button library could have been helpful for detecting when a click happens.

Adjustment from Arduino Uno to Adafruit Feather

When designing the controller glove, we had a goal of making the remote controller as small as possible. At first, we focused on testing with the basic Arduino Uno, but we were unsatisfied with its size. The Arduino Uno did not fit nicely with the glove, and the wires were hard to manage. With multiple testings and adjustments, we were able to switch from the Arduino Uno to the more compact Adafruit Feather, which also made organizing wires easier and prevented them from slipping and falling off the breadboard. At last, our delicate physical product, being the lightest and smallest remote controller in the class, performed perfectly with everything connected securely. With its own 3.3V battery, the glove controller was on its own, travelled freely with the hand.