Hi, my name is Samuel Dauk and I am a Computer Science major with a minor in Applied Mathematics. I enrolled in this class because I have started getting into 3d printing as part of my diy robotics hobby and thought this class would be helpful in learning how to design and create the 3d models.
Project Link: https://asap.csail.mit.edu/
The project inspiration I have chosen is from the MIT Computational Design and Fabrication Group (CDFG) and stood out to me because of its relation to my interest in robotics. The project focuses on how to create an algorithm that allows robots to be given a set of parts and a final product and be able to build the final product out of the parts without specific instructions on how to do so. The algorithm, called ASAP, decides which parts to put where and the order in which it places them. The idea is to mimic how a human would decide to go about building an object if given the same information. Essentially, ASAP is trying to imitate human intuition. ASAP is implemented using weighted tree search methods that are used to determine the feasibility of possible part placements. Below is an image of the algorithm in action showing the chosen process for building a toy camera.
Overall I thought this project was really interesting and helped broaden my idea of what computational design and fabrication can be used for.
Hi, Samuel! Thank you for sharing this! I find it extremely interesting that this algorithm is trying to mimic human intuition. This is the first ever algorithm I know that tries to mimic humans instead of trying to efficiently calculate and execute the tasks.
Hello Samuel! That project is definitely aligned with your robotics interest! Did you know much about that group before completing this assignment? It certainly expands the possibilities of computational fabrication in an interesting way!
Hey Samuel, this project looks amazing. I was curious if this project would be using any deep learning techniques to improve the algorithm, and found that they are in fact using Graph Neural Networks (GNN). I can now say I’m excited to see how ML/AI advancements can supercharge current computational fabrication techniques and projects like this for the feature. It wouldn’t be far-fetched to expect our cars to drive themselves on roads built by similar deep learning models. Great pick!