Cost-effective 3D printing: support structure discovery using reinforcement learning in 3D simulation

Abstract

The project investigates the integration of reinforcement learning techniques, specifically Proximal Policy Optimization (PPO), into 3D printing design, with a focus on material optimisation and support structure discovery. The research explores the capabilities of RL algorithms in optimising decision-making processes for efficient and sustainable manufacturing practices. Through a series of experiments and analyses using a simulated 3D environment, the study demonstrates the agent's proficiency in completing tasks involving simple structures while highlighting challenges in handling larger and more complex configurations. The findings also highlight the potential of reinforcement learning in improving 3D printing processes, but they also emphasize the need for additional research to address scalability issues, improve policy exploration mechanisms, and incorporate real-world variability for comprehensive applications in sustainable manufacturing.