Autonomous Waste Characterization in Environmentally Conscious Decision Making

Product take-back and reuse is an effective way to reduce the environmental footprint of products. Millions of tons of waste are disposed in landfills in the United States, electronic products being of particular concern. While constituting a small fraction of landfilled waste, electronic components account for a majority of the environmental impact. The major challenge in addressing this issue is that the components are functionally obsolete and in a state where their numbers and type are not known. Even with concerted efforts to solve this problem through better design or collection practices, a major unknown is how much actually falls through the cracks and makes it to landfills. Human sorting and identification is impractical, while automating this process has been difficult because of limitations of algorithms to match human ability to discern objects. Deep learning promises to change this. This paper discusses the use of autonomous systems that can scan unorganized heaps of products to identify and catalog components, particularly electronics. This approach can fill an important gap in our knowledge. This paper discusses the testbeds created by the authors which shows promise in accomplishing this task. The paper also discusses the repercussions of such a study and cataloging on design decision-making as well as environmental legislation.