SBIR Phase I: Robotic System for the Sorting of Construction and Demolition Recyclables

The broader impact/commercial potential of this project will be to change the fundamental economics of the recycling process. Analysis suggests that the system can drive the cost of recycling
 to levels competitive with, or below those of, landfilling, drastically changing the waste landscape. No longer will recycling be driven by government mandate, and grow slowly only because of public education and pressure. Instead, market forces will be aligned with waste diversion. The roadmap begins with attacking Construction & Demolition waste, followed by the entire solid waste stream. Although it is estimated that up to 95% of the waste stream could be recycled, only a third of the 250 million tons that are generated each year in the United States is currently diverted. Greater diversion would provide immense savings in landfill and processing costs, and benefit the environment as well. Tens of millions of pounds of greenhouse gas emissions from virgin material mining may be eliminated, and pollution from landfill waste reduced. The existing sorting process is expensive and unprofitable, requiring human workers to manually sort debris, an extremely dull, dirty, and dangerous profession. This innovation has the potential to eliminate these trade offs between cost and environment. This Small Business Innovation Research Phase I project will create a scalable, integrated robotic system that autonomously sorts Construction & Demolition waste for recycling. This advance in autonomous systems is made possible by a series of innovations in robotics: (1) new safety features in robotic hardware that ensure efficient and 
safe collaboration with human workers, allowing for the system to be deployed with virtually zero retrofitting in existing facilities; (2) new motion planning techniques that allow for trajectories to be generated in real time, customized for the characteristics of the waste, safety, and any uncertainty in individual objects? position;
 (3) modern machine learning techniques that allow the system to classify waste at levels approaching human performance, with a continual training signal obtained via human supervision; and (4) tremendous improvements in both computer vision and robotic manipulation, allowing for previously unseen objects to be modeled and manipulated reliably. These innovations pave the way for a new era in recycling, where waste is sorted cheaply, safely, and reliably on a universal scale.