Efficient Sequential Decision-Making in Design, Manufacturing, and Robotics

All dates for this event occur in the past.

Scott Lab E525
201 W. 19th Avenue Columbus, OH 43210
Columbus, OH 43210
United States

Speaker: Md Ferdous Alam, Mechanical Engineering

Abstract:
Traditional design tasks and manufacturing systems often require a multitude of manual efforts to fabricate sophisticated artifacts with desired performance characteristics. Engineers typically iterate on a first principles-based model to make design decisions and then iterate once more by manufacturing the artifacts to take manufacturability into design consideration. Such manual decision-making is inefficient because it is prone to errors, labor intensive and often fails to discover process-structure-property relationships for novel materials. As robotics is an integral part of modern manufacturing, building autonomous robots with decision-making capabilities is of crucial importance for this application domain. Unfortunately, most of the robots in the manufacturing industries lack such cognitive abilities. We argue that this whole process can be made more efficient by utilizing machine learning (ML) approaches, more specifically by leveraging sequential decision-making, and thus making these robotics tasks, design processes and manufacturing systems autonomous. Such data-driven decision-making has multiple benefits over traditional approaches; 1) machine learning approaches may discover interesting correlations in the data or process-structure-property relationship, 2) ML algorithms are scalable, can work with high dimensional problems and learn in highly nonlinear systems where a model is not available, 3) thousands of man-hour and extensive manual labor can be saved by building autonomous data-driven methods. Due to the sequential nature of the problem, we consider reinforcement learning (RL), a type of machine learning algorithm that can take sequential decisions under uncertainty by interacting with the environment and observing the feedback, to build autonomous manufacturing systems (AMS). Due to the poor sample complexity of traditional RL, we propose to utilize transfer learning in the context of RL for building AMS. We propose several transfer RL algorithms and demonstrate their effectiveness in a novel prototypical autonomous manufacturing robot. Furthermore, we explore how we can generalize such decision-making processes to complex high-dimensional tasks such as robots with unknown models. Finally, we investigate an alternative approach for sequential decision-making in high-dimensional design tasks using transformer-based modern neural network architectures for scalability and representation learning. We propose to learn a probability distribution of the latent space of complex sequential design tasks for critical downstream applications. As this research is one of the earliest works on sequential decision-making in the design and manufacturing domain, we anticipate that this research will open new avenues of research in the broader aspect of AI-assisted design and smart manufacturing. We anticipate that our proposed approaches may be specifically beneficial for complex multi-part artifacts design and manufacturing, micro-nano scale manufacturing systems such as Fin-FET transistor design or semi-conductor manufacturing systems in general, optoelectronic device manufacturing such as distributed bragg deflectors (DBR), bio-additive manufacturing, soft robots and flexible sensor manufacturing and many other critical industries.

Zoom Link (or alternative) - if available
https://proxy.qualtrics.com/proxy/?url=https%3A%2F%2Fosu.zoom.us%2Fj%2F99466240874%3Fpwd%3DbGliTTVVVE12V0xHTlBhVVZSU0RQUT09&token=NRVoaMbEci%2B30MF%2Bwc1RC8sArtXIGn7VOrX81v%2Ba4QQ%3D, Meeting ID: 994 6624 0874 Password: 379365

Committee Members
Dr. David Hoelzle
Dr. Parinaz Naghizadeh
Dr. Michael Groeber
Dr. Kira Barton
Dr. Jieliang Luo

Tag: event