Grand Challenges: Programmable Design of Sustainable, All-Natural Plastic Substitutes
Plastic pollution has become one of the global environmental challenges. The widespread replacement of petrochemical plastics with all-natural substitutes can largely attenuate the input of non-degradable wastes to the natural environment. However, the conventional design of experiments probes a broad range of parameters in a scattershot fashion, so trial-and-error cycles are often needed to optimize the recipe of an all-natural substitute with user-designated properties. Moreover, as the number of all-natural substitutes increases, the input time and cost will be inflated accordingly. Therefore, in this Grant Challenge Team Project, the team aims to tackle the global challenge of plastic pollution by accelerating the development of sustainable, biodegradable, all-natural plastic substitutes via artificial intelligence and robotic technologies. The team will employ machine learning (ML), robot–human teaming, in silico data augmentation, and molecular dynamics (MD) modeling to construct a high-accuracy prediction model. The prediction model can conduct two-way design tasks, including (1) predicting the optical, thermal, and mechanical properties of an all-natural substitute from its designated composition and (2) automating the inverse design of all-natural substitutes with well-matched properties. The team will utilize the integrated workflow to develop and even commercialize multiple sustainable, biodegradable plastic substitutes with user-designated optical, fire, and mechanical properties, the recipes of which were automatically predicted without any trial-and-error cycles.