Most families eventually throw away a soft cucumber or a browning avocado. Thin plastic wrap usually covers that produce, yet it does little to stop mold. Now, engineers have paired a robot with artificial intelligence (AI) to design better food wraps. Their biobased food packaging kept cucumbers and avocados fresher than commercial plastic in laboratory tests. Po-Yen Chen led the research at the University of Maryland, College Park. Yang Li, Tianle Chen and Peihua Ma share first authorship. In addition, partners came from USDA’s Agricultural Research Service, Iowa State University and Yale University. The team published the study in Science Advances on October 28, 2026. Readers can find it through DOI 10.1126/sciadv.aef6378 (Li et al., 2026). As a result, anyone can now explore the data on a public platform. The project also connects chemistry, food science, robotics, and machine learning in one ambitious effort.
Challenges and Motivation Behind Biobased Food Packaging
Grocery packaging depends on petroleum plastics such as polyethylene and polyvinyl chloride, or PVC. Manufacturing them consumes considerable energy and releases greenhouse gases. For example, American PVC production emitted about 18 million metric tons of CO2 in 2020. Afterward, most plastic waste, over 85%, accumulates in landfills and oceans. There, it can release tiny fragments called microplastics. On top of that, ordinary wrap has no built-in defense against microbes. Because of this, fruit and vegetables can still spoil inside it. For this reason, the authors see antimicrobial packaging as a way to extend shelf life. Spoiled produce becomes food waste, which the United States wants to reduce. Indeed, the U.S. Environmental Protection Agency (n.d.) aims to halve food waste by 2030. Packaging that preserves produce longer could therefore support that national target. Still, any replacement material must remain transparent, durable and safe for food contact.
Billions of Possible Biobased Food Packaging Recipes
Natural materials offer a greener path for food packaging. Cellulose, gelatin, starches and clays are plentiful and renewable. Many also decompose in soil, a goal shared by other biodegradable plastics. However, choosing the right combination is extremely difficult. The Maryland team assembled a materials library of 23 natural or food-safe ingredients. Overall, the library contained 2 nanofillers, 13 biopolymers and 8 additives. Next, the researchers calculated the size of the design challenge. Selecting any five ingredients in 2% steps produces more than 10 billion recipes. Fabricating and testing every recipe would take over 1.8 million years. In other words, manual trial and error could never cover that territory. At the same time, each film must satisfy several requirements together. For instance, it should be transparent, mechanically strong and environmentally friendly. As a result, the team chose a faster, data-driven strategy instead.
Science Explained: How a Robot and AI Design Films
Think of the workflow as an intelligent kitchen that writes its own recipes. First, an automated pipetting robot combined liquid ingredients in many different ratios. The mixtures then dried into thin experimental films. Altogether, the robot formulated more than 2,420 films for this first screening stage. Researchers then checked whether each film peeled off cleanly and looked smooth and uniform. Those feasible-or-infeasible results then trained a group of five neural networks. A neural network is a computer program that learns patterns from examples. As a result, the model learned which recipes would likely form a usable film. In other words, this step mapped the workable region before any strength testing began. Similar AI assistants now guide other fields, as AI agents for inorganic materials show. Together, robot and software turned guesswork into a continuous cycle of making, measuring and learning.
Fig. 1: How the Robot-AI Loop Works
Active Learning and Smarter Chemistry
Next came active learning, where the model selects its own upcoming experiments. In each loop, it picked uncertain or especially promising recipes. The robot prepared those films, and the team measured their clarity and strength. Over 16 loops, the laboratory fabricated and tested 343 films. By the end, prediction error on independent test data dropped to about 16%. The final model then predicted properties for approximately 1 billion candidate recipes. At the same time, the team applied density functional theory, or DFT. This computational method estimates how strongly atoms and molecules bind together. It showed that calcium ions bind most strongly inside cellulose-rich films. In turn, calcium cross-linking raised film strength from 220 to 320 megapascals, peaking at 339. Another model-selected film reached 151 megapascals while staying 90% transparent. In addition, bags made from three suggested films each carried over 3 kilograms of produce.
Key Findings From Biobased Food Packaging Tests on Produce
Strong films matter only if they protect real food. So the team added copper ions to a chitosan-rich film. Chitosan is a natural polymer extracted from chitin. Copper fights microbes and also helps the film resist water. In laboratory tests, liquid soaked from this film largely suppressed E. coli growth. Then the researchers wrapped cucumbers and refrigerated them at 4 °C for 15 days. Each storage condition included 20 separate cucumbers. Unwrapped cucumbers showed decay, softening and extensive mold growth. By contrast, copper-chitosan cucumbers displayed minimal mold. Electrolyte leakage, a sign of damaged cells, reached 54% in unwrapped cucumbers. It measured 34% under commercial LDPE wrap and just 21% under the new film. In addition, the film delivered the lowest water loss and yeast and mold counts. The study (Li et al., 2026) credits germ control, moisture control and an oxygen barrier.
Avocados Stay Firm Under a Cellulose Film
Avocados required a different formulation. Here, the model recommended a film rich in cellulose nanofibers, tiny threads from plant material. Once again, copper provided germ protection as well as water resistance. The finished film transmitted about 91.8% of visible light. Next, the team wrapped Hass avocados and stored them at room temperature for 7 days. The researchers tested six avocados for each condition. Unwrapped fruit developed blackened skin and browning pulp. Likewise, LDPE-wrapped avocados darkened extensively. In contrast to that, avocados in the copper-cellulose film kept green skin. Firmness indicates how ripe and fresh an avocado remains. Under the new film, firmness reached 587.7 gram-force after 7 days. At the same time, LDPE fruit measured 201.3 gram-force, and unwrapped fruit only 118.3. As well as that, the film reduced mass loss and microbial counts. These results come from laboratory storage, not grocery shelves.
Copper, Cell Safety and Soil Checks
Because copper fights germs, the team checked whether any reached the food. After 15 days, cucumber pulp contained about 0.4 parts per million of copper. That level matched unwrapped cucumbers, so migration was negligible under these conditions. In the same way, avocado pulp showed no copper above control levels after 7 days. The researchers also performed cell safety tests on liquid soaked from the films. Mouse fibroblast cells remained healthy at film amounts up to 200 micrograms per milliliter. Twelve additional films passed a second cell test as well. After that, the team buried several films in soil beside LDPE and HDPE plastics. Within 60 days, every biobased film visibly disintegrated. On the other hand, both plastics stayed intact. Even so, this was a qualitative screening test, not a formal compost certification. Estimated soil copper also remained far below US EPA screening levels.
Fig. 2: Key Takeaways From the Packaging Study
Implications: Business, Careers and Limits
This study shows how AI can accelerate the search for greener materials. Yet laboratory speed is only the first step. The authors performed a life cycle assessment of environmental impacts. It tallied impacts from raw materials to the finished film. That assessment stops at the factory gate, so disposal and composting remain unexamined. The chitosan film outperformed plastics in several categories but lagged in others. For example, ozone depletion scored worse, partly due to chemical-intensive chitosan extraction. In response, the model searched again and found a chitosan-free cellulose recipe. That film beat conventional plastics in nearly every category. For now, the team produced films only through laboratory-scale blade coating. The authors say this suggests compatibility with roll-to-roll processing, though it remains untested. They also call for a full cost analysis. In the meantime, startups can explore recipes on the team’s public data platform.
Expert Commentary From the Research Team
An engineering news release (A. James Clark School of Engineering, 2026) shares the senior author’s perspective. In it, Po-Yen Chen explains the project’s main goal. “Food packaging needs to do more than replace plastic,” Chen said. He added that packaging “also needs to protect the food inside.” That idea matches the produce trials, which measured freshness rather than simple plastic swaps. Chen also described the value of automation. With AI and robots combined, he said, “we can accelerate the development of bio-based materials.” In his words, those materials could “help keep produce fresh and reduce food waste.” After all, a human team could never test billions of combinations by hand. Instead, the robot and model narrow the list to a few strong candidates. Even so, faster development does not equal a finished product. Ultimately, scale-up and cost will decide whether stores adopt these films.
Careers in Biobased Food Packaging
This project shows how one packaging problem connects many professions. Chemists select the ingredients and the cross-linking ions. Meanwhile, food scientists measure freshness, mold, firmness and safety. Robotics engineers build systems that mix and cast hundreds of films. Data scientists train models that predict which recipes will succeed. As a result, teens who enjoy chemistry, cooking, coding or robots all belong here. Useful college majors include chemical engineering, food science and materials science. Computer science and data science, in turn, suit the machine-learning side. In the meantime, high school students can strengthen chemistry, biology and mathematics. Learning Python also helps with laboratory data. For example, lab automation engineers keep pipetting robots like this one running. Related ideas, such as programmable plastics that break down on cue, hint at future careers. Factories will also need packaging engineers and quality testers whenever new films scale up.
Conclusion: Faster Discovery, Real Tests Ahead
Robots and AI helped University of Maryland engineers design biobased food packaging remarkably quickly. The project began with just 23 natural ingredients and billions of possible recipes. Through active learning, the team trained a model that scanned about 1 billion recipes. Then DFT pointed to calcium and copper ions for tougher, more water-resistant films. During testing, cucumbers and avocados stayed fresher than under commercial plastic wrap. Copper remained at control levels in the produce, and the films disintegrated in soil. However, the work remains laboratory-scale, and no cost analysis exists yet. The life cycle study also stops at the factory gate. For these reasons, factory trials and complete cost data are the next major steps. Even so, the open data-sharing platform lets students and companies test ideas today. In the long run, smarter wraps could mean less waste in every kitchen.
The key numbers match the text: 2,420 screening films and 343 active-learning films, about 16% prediction error, 320 ± 19 MPa tensile strength, cucumber electrolyte leakage of 21% versus 34% for LDPE and 54% for unwrapped, and avocado firmness of 587.7 versus 201.3 gf. Three overstatements were corrected: a “cytotoxicity check” and named industrial coating methods are not in the retrieved text and were removed, and copper was measured only in cucumber flesh, not all fruit. The “design tool” is now called a public data platform. The license variant remains unconfirmed. Verdict: accurate after corrections.
ENTECH runs this story because it joins two fast-growing fields, sustainable packaging and AI-driven “self-driving” labs, and tests the result on real food. University of Maryland and USDA researchers showed that robot-made biobased films can match or beat plastic wrap for keeping cucumbers and avocados fresh, and they shared their data on a public platform. For teens, it links chemistry, food science, robotics and machine learning in one project. The caveats: this is lab-scale work, cost has not been analyzed, and the environmental assessment stops at the factory gate. Takeaway: AI can speed up the hunt for greener materials, but scale-up and cost will decide adoption.
Sailee Pakhare holds a Master's degree in Biotechnology, with a research background spanning microbiology, molecular biology, and disease biology. Her postgraduate dissertation examined the antimicrobial potential of bioactive peptides from lactic acid bacteria, and her research experience includes molecular identification of bacterial isolates (with sequencing data submitted to NCBI) and work as a Research Assistant at SafeWright Laboratory. She also completed a summer internship at IIT Hyderabad studying circadian rhythms and disease biology using Drosophila melanogaster as a model organism.
Her work has been recognized with the Best Poster award at the 3rd International Conference on Innovations in Biotechnology Research for Sustainable Bioresources and Bioeconomy (2025), and she has contributed to a book chapter on the role of AI and machine learning in climate change.
Sailee is also a co-founder of LifeXtension, a community focused on longevity, ageing, and health science awareness in India. An interdisciplinary thinker, she explores questions ranging from disease biology to space and the origins of life, approaching science as a way of asking questions rather than a fixed body of facts. As a science communicator and researcher, she is committed to translating complex research into writing that is clear, accurate, and engaging.
