Elopement can be a looming, terrifying problem for parents and caretakers of neurodiverse children (e.g. those with autism). These very bright, goal-oriented children decide to leave the care of family and staff either in pursuit of or to get away from something, seemingly without notice or plan.
According to Seattle Children's, studies have found that 49% of children with autism age 5 and older eloped, and 53% of those were gone long enough to be considered missing.
The Idea
Project Kong is designed to be a small, low-cost, completely autonomous rover to track eloping children and notify caregivers of the location of eloping children in their care. Leveraging the Donkey Car platform keeps the cost down, while a few key upgrades makes it easier to train the Car on the field and adds the capability for tracking over BLE.
The Process
Rather than lean on prose and images, I decided to record a video series documenting the full build, testing, and demonstration. The full video series is included below save the demonstration included above.
Updates to the Donkey Car
While the Donkey Car provides a capable platform for basic autonomous driving via neural network, the Donkey Car alone proved lacking in a few areas key to this project:
I [added a new part](https://github.com/wroscoe/donkey/pull/335) to allow control of the Raspberry Pi (e.g. during training) over the RX/TX pins, which I used to drive the car with the original RC controller.
I [added a failsafe](https://github.com/wroscoe/donkey/pull/329) for loose connections from the Raspberry Pi. Some of the wires in the original kit had loose connectors, so sending PWM impulses from the Pi would crash the software.
I've already shown Kong as a shining example of how AI can be used to support the marginalized in our society, rather than just the pockets of the privileged. With Arm's support and the Maker spirit, Kong came together as an effective proof-of-concept that AI can be used effectively and safely in the home in novel ways.
Comments