Reducing Computational Time in Pixel-Based Path Planning for GMA-DED by Using Multi-Armed Bandit Reinforcement Learning Algorithm

This work presents an artificial intelligence technique to minimise path planning computer processing time for successful GMA-DED 3D printings. An advanced version of the Pixel space-filling-based strategy family is proposed and developed, using, originally for GMA-DED, an artificially intelligent Reinforcement Learning technique to optimise its heuristics. The initial concept was to boost the preceding Enhanced-Pixel version of the Pixel planning strategy by applying the solution of the Multi-Armed Bandit problem in the algorithms. Computational validation was initially performed to evaluate Advanced-Pixel improvements systematically and comparatively with the Enhanced-Pixel strategy. A testbed was set up to compare experimentally the performance of both algorithm versions. The results showed that the reduced processing time reached with the Advanced-Pixel strategy did not affect the performance gains of the Pixel strategy. A larger build was printed as a case study to conclude the study. The results outstand the artificially intelligent role of the Reinforcement Learning technique in printing more efficiently functional structures.