Build Process

The first step in the build process was to decide the components in the receiver and wand as well as how they would communicate. Our requirements called for the receiver to be encased in resin and for LEDs to light up in response to different predefined user gestures.

We decided upon using a Raspberry Pi 4b as the processing unit. This was determined to be necessary over other microcontrollers such as ESP32s because of the processing power required for gesture/image recognition software that we planned to use, as well as its ability to easily interface with high-resolution cameras. The wand was determined to emit an IR signal for tracking by the camera in the receiver as well as include an inertial measurement unit for more complex gestures. Wireless charging for the receiver was chosen to prevent the need for cables poking out of the cube.

The first IR test was done with the camera and an IR filter in front of it. The IR LED is attached to a battery pack, forming our first prototype wand. The LED is clearly visible and software was written to trace the path of the wand, which can be seen here. Eventually we would use this basic tracing function to be able to recognize different gestures made by the user.

The most challenging part of this was ensuring that the camera remained solely focused on the IR LED and would not attempt to track other features in the environment. Although we have mostly been able to mitigate this as an issue, it can sometimes occur under unique lighting circumstances.

In the initial design and testing phase, the team was worried about how heat dissapation would be affected by the electronics being embedded in resin. In order to determine the thermal properties of the resin, we tested the power limits of it. Using a 1Ω equivalent ciruit, the resin ended up cracking at 15W of power. This set the upper threshold for power consumption of our system to build around.

All components that would be placed in the resin cube were individually tested in resin before the first prototype cube was assembled. The LED strip seen above was one of the components. Although we assumed certain components such as the LEDs would behave normally, we wanted to eliminate risks that would otherwise be hard to find when the entire system was assembled.

The first resin cube can be seen on the left. It was set in a 3D printed box as its mold and have few functional components. The main goal of this cube was to analyze the full setting process as well as peak temperature during the set. The ratio of epoxy resin to the hardening agent is crucial. We had not measured them precisely enough, resulting in the resin taking multiple days to fully set and a sticky texture on the outside.

The cube's maximum temperature was 70°C and was affected by the amount of resin setting at a given time. Additionally, the 3D printed cube was not an optimal mold as the cube was extremely difficult to remove. As a result, we began to create our own silicone molds that could be shaped to specific sizes and were far easier to remove the cubes from. One of the molds can be seen on the right.

The second resin cube developed can be seen on the left. It included a functional Raspberry Pi 4b and LED PCB on the front side. The Raspberry Pi and LEDs were able to light up through an external power from outside the cube. However, the elements in the cube would shift while it was setting, resulting in a messy look inside of the resin. The resin was also not as clear as we wanted it to be.

We preformed multiple tests to determine the cause of the resin not setting in a clear manner and concluded that it was due to air bubbles appearing when it was stirred too quickly. Additionally, a 3D printed stand was printed for the components to rest on so that they do not move when the resin is being set. The first stand prototype can be see on the right.

The cube seen on the left was the next iteration of the cube that included significant improvements. The battery and charging system worked inside the cube and the Raspberry Pi's camera was fully functional, albeit slightly distorted.

The cube on the right is considered out minimum viable product. The resin has been properly set and the electronics are fully functional. Although the cube works, the gesture recognition was not fully accurate due to the positioning of the camera being slightly further back in the cube and some air bubbles in front of it distorting the tracking capabilities.

The wand was more difficult to develop than we had initially imagined it to be. The circuit is simple, with only a battery, a button to turn it on, and an IR LED. The team first considered making it out of resin, like the cube, but encountered difficulties with this approach.

By encasing it in resin, it was difficult to get the components set correctly and made switching the battery in it difficult. The button also frequently got stuck, rendering it unusable. The team switched to 3D printed cases for the wands, with the electronics mounted inside. This proved more reliable and easier to produce, making it the optimal solution for us.