ML-Assisted Autonomous Assistant Robot for High-Throughput Algal Bioplastics Fabrication

To meet the pressing global demand for reducing the harmful effects associated with the sourcing, manufacturing, and disposing of synthetic polymeric materials and to achieve a more circular global economy, we aim to create a new family of fully bio-based plastic alternatives that can easily degradable in natural environments with minimal waste and energy expenditure. In previous work, we have devised a method to fabricate strong and stiff bioplastic materials directly from entire cells or tissues (biomatter) through heat and pressure. However, due to the complex nature of little-to-unrefined biomatter feedstock, the most significant limitation in our ability to design and optimize this family of bioplastics is the difficulty in understanding the fundamental mechanisms that control the transformation from biomatter feedstock to a cohesive bioplastic and deconvoluting the factors that contribute directly to the final material properties.
Our ongoing effort to gain a more comprehensive understanding of bioplastics, in particular those made using algal biomatter feedstock, uses machine learning (ML) models to establish a framework that correlates processing conditions of algal bioplastics to their material properties. In order for the ML model to produce adequate results, a large dataset is essential; hence, there is a need to efficiently and reliably fabricate a large number of samples for characterizations and testing.
This work aims to increase the rate of sample fabrication via automation and robotic assistance, hence eliminating the constraints to production rate associated with human factors in our existing workflow. In the current stage, we plan to implement automation in the material dosing, mixing, preparation of molds for compression molding, and certain sample characterizations. We also aim to have sample characterization results compiled and uploaded to the ML algorithms automatically and have the ML outcome to then set the parameters for the next experiment. This system will create the infrastructure required to establish the knowledge framework for deconvoluting the processing-property relationships in algal bioplastics, which we hope can be universally adapted for bioplastics using other feedstock sources.