Are you ready for your internship at ToThePoint?
TrashBeat – Build a smart trashcan
Design and develop a self-learning trashcan, by using sensors and a backend processing pipeline with machine learning, to create an intelligent and timely signal when it’s time to be emptied.
Description of the assignment
- In our ToThePoint offices of the Cronos campus, there are a lot of trash cans to keep everything tidy.
- We want to add sensors to the trashcans to collect data so an intelligent backend service can signal in time when a certain trash can should ideally be emptied.
- The ideal time to send out the signal is not necessarily when a certain capacity is exceeded. Because for example: there is a difference between a trashcan that has only 30% capacity left on Wednesday afternoon at 1PM or the same trashcan on Friday afternoon at 1PM – because there are different usage scenario’s in place on both of these days.
- One way to map the usage patterns is to define it in a deterministic manner. But to us it seems more learnative (and more correct) to let the trashcans themselves learn to become aware of their environment and the usage patterns surrounding it.
Your assignment is therefore twofold:
- Realize a hardware component (select and implement sensors and its microcontrollers) so you can start measuring and collecting data to send it to the processing backend
- Develop software to recognize patterns in the usage data and have it send out notifications based on these patterns so the trashcans can be emptied in a smart point in time.
Goals
- Develop an IoT architecture with attention for bot hardware and software (such as energy consumption and scalability of the processing software-side)
- Develop a hardware prototype
- Develop a backend business logic
- Iteratively build a machine learning model to make smart calls about when it is or isn’t
necessary to initiate an emptying of the trashcan - Develop a visualization of progress
What you will gain
- You’ll learn to prototype and adjust your product
- Capture and process relevant real-time data
- Explore the possibilities and limitations of sensors
- Gain knowledge and experience with designing and deploying a machine learning solution
- That lovely feeling you’ll get knowing your design will effectively be used in a real-life scenario
What you need
- Creativity and the will to succeed
- A motivated personality to keep pushing until you find the right solution
- You’d love to see hardware and software working together
- You love to explore real-time data and stream processing
- You acknowledge that an exciting time is ahead with machine learning applications
- You’re looking forward to learning a heck of a lot in a relatively short time period
Technologies you'll be using
- Microcontrollers
- Cloud backend
- Messaging infrastructure
- Spark and MLLib
- ReactJS and D3.js
Location of your assignment
Veldkant 33B, 2550 Kontich
Your mentor
Kevin Smeyers – Technical lead machine learning ToThePoint
