The project is motivated by the need of small-to-medium sized businesses in the United States to automate handling and assembly of compliant parts. As a case study, it focuses on robotic installation of a wire harness (a bundle of wires that terminate in electrical connectors). This manufacturing task is ubiquitous but hard to automate because it requires reasoning about deformation of the wire harness. The project is addressing this challenge with new algorithms for manipulation and perception of deformable objects that are based on novel representations of object shape. It considers first the manipulation of a single wire (one piece of a wire harness). It expresses the shape of this wire as the solution to a geometric optimal control problem, and shows that the set of all solutions to this problem is a smooth manifold that can be parameterized by a single coordinate chart. This result leads to an algorithm for manipulation planning that works well and is easy to implement. Objectives include extending this model to consider variable stiffness, plastic deformation, and branching; making manipulation plans robust to perturbation; estimating material properties and shape from sensor data; and experimentally demonstrating robotic installation of a wire harness. Outreach efforts include co-directing a week-long summer institute for high school students, mentoring undergraduate researchers, and organizing an industrial forum.
