Building a smart cigarette cleaning robot in 2 weeks
how angsa robotics came to be
One of the many reasons I decided to go to study abroad is to learn how the big startups work (and I was really lucky to get a part-time opportunity at FlixBus), and also how creative minds solve problems and may be involved in building another startup.
That’s why when I heard about Think.Make.Start I instantly want to participate in the course (yes, this is an awesome university course that I highly recommend everyone to do). The course itself is giving you lots of freedom. You’re supposed to come up with problem(s), gather like-minded minds and build a working prototype in 2 weeks. And sometimes the idea and/or the prototype will be developed further into a viable business.
We often heard that the early stage of startups contains several key persons: the hipster, the hustler, and the hacker. But to be honest, this time I kinda want to position myself as the observer/supporter. That’s why instead of actively finding problems and looking for potential teammates, I just wander around asking other people and try to dig deeper into their problems, concepts, and methodologies. I observed some common topics occurring: healthcare/related tech (I mean, it’s happening all over the globe right now), education tech, and sustainability tech. I was into the healthcare and sustainability tech and decided to focus on the individual/potential team exploring those ideas.
Some of the ideas struck me as awesome, some individuals impressed me with their vast knowledge and skills but somehow none of those blends well together in my limited opinion/observation. But since the team-seeking-building process has a time limit, I aligned myself with some individuals with the most interesting problems in the sustainability aspect. Luckily those people also end up making a team together.
After that we do the usual idea building journey:
- Figuring out what exactly we want to build. We iterate with some ideas in mind, the feasibility and possibility aspect of the product and other kinds of stuff. Finally, we decided to build a smart (with computer vision and simple neural network) robot to clean small trashes. What’s with the fancy smart stuff you might ask? Well, we need to be able to differentiate trash (like cigarette butt) with other stuff that we’re not supposed to clean (like bugs in the park). Imagine like a Roomba for outdoor trash. It was originally Karl’s idea with small adjustments added during our discussions.
Fun fact: Angsa is an Indonesian word for swan since our prototype was somehow resembling swan so we decided to use angsa as the name instead of c̶i̶g̶a̶r̶e̶t̶b̶u̶t̶t̶s̶u̶c̶k̶e̶r̶3̶0̶0̶0̶.
- Iterating the prototypes. This is where Think.Make.Start gets super interesting. They provide us with hardware libraries that we can borrow, reasonable budgets that we can spend and mentors that are awesome. We divide the prototype building into several tasks:
- Trash picking mechanism: we tried some different mechanisms such as picking with a needle, multiple needles, with different angles until we settle on using a modified vacuum cleaner as it will provide the solution for trash picking and storing at the same time.
- Movement: we used Arduino board to control 2 motors and the vacuum cleaner. Originally we plan to add ultrasonic sensors as obstacle detectors but unfortunately, there was not enough time to make this work.
- Trash detector: for the first MVP, we decided to use simple computer vision to detect cigarette butts because using neural network is simply not feasible at this stage (no model, not enough train data, etc) and we use Raspberry Pi to do this task as well as sending command to Arduino. For the first MVP we decided to focus on cigarette butt detection because it’s super dangerous.
- Finding potential clients. Things are getting harder at this stage. Cold calling people (with language that I’m not familiar with a.k.a Deutsch), meeting and discussing possibilities and other business-y stuff while still perfecting the prototype. I was really surprised to know that most of the businesses that we contacted were supportive and some of them even willing to sign Letter of Intent even though we haven’t got any working prototype yet.
- Demo day. We won THINK award! There are several winner categories and THINK award is rewarded to the idea with the most support from the audience.
After Think.Make.Start we discuss briefly on the future of the prototype. Karl and Lukas are strongly interested and want to approach this as a full business idea, while the others (me, Daniel & Denim) even though interested as well can’t really put our full-time effort on this product.
Since I do have my other activities like academic courses, work at FlixBus and Medigo Indonesia I couldn’t help that much in the 2nd prototype iteration. In this iteration, we try to improve processing power by using Nvidia Jetson for trash detection. We also start to create a neural network by collecting the input data, labeling it and creating a model to detect trashes. This is the preview of the 2nd prototype presented in the Green Challenge event in Denmark. Even though we didn’t win, we get some support and even some pilot offers.
Now, Angsa is under continuous technical/business development and I do have high hopes for the future of this product. Aside from the noble goal of making a sustainable future, this will also show the possibility of integrating artificial intelligence/computer vision further into our daily life. If you’re interested, want to contribute, or have some interesting offer feel free the check website below ;)
Startseite - Angsa Robotics
Angsa entwickelt ein künstliches neuronales Netzwerk, das in der Lage ist, selbst kleine und teilweise verdeckte…