The UW + Amazon Science Hub has announced its 2026-27 Robotics Research Awards. The awards invest in promising early-stage robotics and AI research, helping UW faculty turn ambitious ideas into future breakthroughs. The following six projects were selected for the UW + Amazon Science Hub Faculty Research Awards and will each receive $150,000 in one-year gift funding from Amazon.
Coupling Curl, Kinks & Flexible Vitrimers to Engineer Flexible Paper Packaging

PI: Aniruddh Vashisth, Assistant Professor, Mechanical Engineering
The goal of this project is to create a recyclable paper packaging material that performs more like plastic. The team is engineering a new type of paper-based composite that can stretch, absorb shocks, and withstand the demands of shipping and handling. The material can also be repaired and reused after damage, extending its lifespan and potentially reducing waste. The technology could offer Amazon and other companies a more sustainable replacement for traditional plastic mailers used in e-commerce.
Recovering Extensibility in Recycled Paper with Low-Loading Supramolecular Biopolymer Modifiers

PI: Eleftheria Roumeli, Associate Professor, Materials Science & Engineering
The goal of this project is to create stronger, more flexible paper packaging that can replace materials that are harder to recycle. Researchers are studying how natural, plant-based additives can improve recycled paper, making it more durable, stretchable, and better at absorbing impacts. The project aims to develop recyclable packaging materials that work with existing manufacturing and recycling systems to help reduce waste and improving performance.
Foundation-Model-Guided Risk-Aware Coordination for Large-Scale Robot Fleets

PI: Lillian Ratliff, Associate Professor, Afroze and Sherali Dhanani Endowed Faculty Fellow, Electrical & Computer Engineering

Co-PI: Xuegang (Jeff) Ban, William M. and Marilyn M. Conner Endowed Professor, Civil & Environmental Engineering
The goal of this project is to help large fleets of robots work together more efficiently by reducing traffic jams and delays. The researchers are developing an AI-powered coordination system that predicts congestion, adapts to changing conditions, and helps robots choose the best tasks and routes in real time. By preventing small bottlenecks from becoming larger disruptions, the project aims to improve the speed, reliability, and overall efficiency of warehouse operations.
Just enough physics: Approximate geometry for contact-rich reactive manipulation in densely packed bins

PI: Joshua Smith, Milton and Delia Zeutschel Professor, Electrical & Computer Engineering, Computer Science & Engineering
The goal of this project is to help robots handle items in bins more accurately and efficiently. The researchers are developing a system that allows robots to better understand how objects are arranged and how they will move when pushed, lifted, or separated. By continuously using sensor data to update its understanding of the scene and adjust its actions, the robot can make better decisions even when information is incomplete or imperfect.
RoboMolmo: A Foundation Model for Whole-Body Long-Horizon Mobile Manipulation

PI: Ranjay Krishna, Computer Science & Engineering
The goal of this project is to build more capable and adaptable robots that can carry out long, complex tasks from natural language instructions. RoboMolmo combines advanced AI reasoning with robot control, allowing robots to interpret a request, plan a sequence of actions, and coordinate movement and object manipulation. By training the system in extensive virtual environments, the team aims to create robots that can generalize to a wide range of real-world situations and perform new tasks with minimal training.
Prediction-Augmented Network Coordination for Robot Fleet Congestion

PI: Jing Yu, Assistant Professor, Electrical & Computer Engineering

Co-PI: Ashis Banerjee, Associate Professor, Industrial & Systems Engineering,
Mechanical Engineering
The goal of this project is to help large fleets of robots move goods more efficiently by reducing congestion and unexpected delays in warehouses. The project will develop a system that uses traffic predictions to coordinate robot movements and task assignments, while accounting for the fact that forecasts are not always perfect. By helping robots be more robust to prediction uncertainty, the project aims to improve warehouse speed, efficiency, and reliability while preventing small prediction errors cascading into warehouse-wide congestion.