Inspiration

We were inspired by different survival/team-based video games, where communication was a neccesity to survive. We thought about a scenario where the world has crumbled, leaving us without Wi-Fi or Cellular connection. Thus, we came up with the idea of a completely off-grid pager to communicate to fellow survivors via text.

What it does

The device accepts text-based input to write out a message, which is then encrypted and sent over radio frequencies. Another device will be in receive mode searching for incoming messages, and upon receival it will decrypt the message and display it on it's on-device screen. This protocol requires no Wi-Fi or cellular connection. Additionally, the pager has a built in camera which is loaded with a custom trained AI model to detect items such as food and wood. Upon scanning an object, it pages other people about what you found.

How we built it

Hardware: ESP32, ESP32-CAM, OLED Display, push button Software: C++, Arduino libraries, Edge Impulse for model training

Challenges we ran into

We originally had our AI model training for image classification as we were going to send a picture, however due to the extremely low quality camera and varying lighting conditions, it performed poorly. We changed our approach to object detection and transitioned from taking pictures to a recording/scanning approach. Additionally we trained it on our own pictures rather than public datasets, as it better represented the conditions we were testing in.

Additionally, since we used the ESP32-CAM rather than a standalone camera peripheral, it provided issues because it also is a MCU, making it difficult to implement as a standalone camera without very specific hardware. We leverage this by actually putting our model onto the camera, lifting some of the computational weight off of the main board, and allowing our code to be better organized.

Accomplishments that we're proud of

Learning to interface a custom trained Edge Impulse model with ESP32 hardware. Implementing wireless connectivity between ESP32's without use of Wi-Fi or bluetooth.

What we learned

How to create a model using Edge Impulse, and collecting data ourself. Comunicating between ESP32's with encryption.

What's next for Grid-Down Pager

Drastically improving the AI model. Finishing the game/simulation environment.

Built With

  • c/c++
  • edgeimpulse
  • esp32
  • platformio
Share this project:

Updates