Web - Sign Up
Web - Store
Web - Confirmation
Android - Face Scan
Hardware - Neighborhood
Hardware - Early
Hardware - Nearly Complete
DeliverAI was inspired by the current shift we are seeing in the automotive and delivery industries. Driver-less cars are slowly but surely entering the space, and we thought driverless delivery vehicles would be a very interesting topic for our project. While drones are set to deliver packages in the near future, heavier packages would be much more fit for a ground base vehicle.
What it does
DeliverAI has three primary components. The physical prototype is a reconfigured RC car that was hacked together with a raspberry pi and a whole lot of motors, breadboards and resistors. Atop this monstrosity rides the package to be delivered in a cardboard "safe", along with a front facing camera (in an Android smartphone) to scan the faces of customers.
The journey begins on the web application, at link. To sign up, a user submits webcam photos of themselves for authentication when their package arrives. They then select a parcel from the shop, and await its arrival. This alerts the car that a delivery is ready to begin. The car proceeds to travel to the address of the customer. Upon arrival, the car will text the customer to notify them that their package has arrived. The customer must then come to the bot, and look into the camera on its front. If the face of the customer matches the face saved to the purchasing account, the car notifies the customer and opens the safe.
How we built it
As mentioned prior, DeliverAI has three primary components, the car hardware, the android application and the web application.
The hardware is built from a "repurposed" remote control car. It is wired to a raspberry pi which has various python programs checking our firebase database for changes. The pi is also wired to the safe, which opens when a certain value is changed on the database.
_ note:_ a micro city was built using old cardboard boxes to service the demo.
The onboard android device is the brain of the car. It texts customers through Twilio, scans users faces, and authorizes the 'safe' to open. Facial recognition is done using the Kairos API.
While not included as a primary component, Firebase was essential in the construction of DeliverAI. The real-time database, by Firebase, is used for the communication between the three components mentioned above.
Challenges we ran into
Connecting Firebase to the Raspberry Pi proved more difficult than expected. A custom listener was eventually implemented that checks for changes in the database every 2 seconds.
Calibrating the motors was another challenge. The amount of power
Sending information from the web application to the Kairos API also proved to be a large learning curve.
Accomplishments that we're proud of
We are extremely proud that we managed to get a fully functional delivery system in the allotted time. The most exciting moment for us was when we managed to get our 'safe' to open for the first time when a valid face was exposed to the camera. That was the moment we realized that everything was starting to come together.
What we learned
We learned a ton. None of us have much experience with hardware, so working with a Raspberry Pi and RC Car was both stressful and incredibly rewarding.
We also learned how difficult it can be to synchronize data across so many different components of a project, but were extremely happy with how Firebase managed this.
What's next for DeliverAI
Originally, the concept for DeliverAI involved, well, some AI. Moving forward, we hope to create a more dynamic path finding algorithm when going to a certain address. The goal is that eventually a real world equivalent to this could be implemented that could learn the roads and find the best way to deliver packages to customers on land.
Problems it could solve
Delivery Workers stealing packages or taking home packages and marking them as delivered. Drones can only deliver in good weather conditions, while cars can function in all weather conditions. Potentially more efficient in delivering goods than humans/other methods of delivery