Robotic handling of flexible objects could lead to streamlined manufacturing
Many small- to medium-sized businesses are currently in “no-man’s land” when it comes to automating the manufacture of compliant parts, as they don’t produce enough volume or capital to justify the use of traditional industrial robots or approaches to automation. This often requires them to use manual labor to craft their products.
CSL Associate Professor of Aerospace Engineering Timothy Bretl is working to develop a manufacturing solution for this target market and recently received a three-year, $450,000 National Science Foundation grant to do research on manipulating deformable objects for robotic manufacturing.
“It’s very simple from a human’s perspective, but for a robot, these things are very difficult. As soon as you go from rigid to flexible objects, there’s a lot of complexity that has been very hard to deal with in the past,” Bretl said.
Bretl is working with graduate students Andy Borum and Dennis Matthews to develop a unique solution to this problem, the foundations of which have been studied for hundreds of years. Borum is primarily focusing on the theoretical side of the research, while Matthews tackles the experimental work.
To accomplish their goal, Bretl and his team are pushing the boundaries of a mathematical theory called optimal control. A typical application in the past might have been to plan the trajectory of a spacecraft flying from the Earth to Mars while consuming the minimum amount of fuel. However, instead of looking for the optimal trajectory between a given start and goal, Bretl and his team are looking for the entire set of optimal trajectories over all possible starts and goals. With the right choice of optimal control problem, elements of this set correspond to shapes of a deformable object.
“What we’re interested in is the set of all possible solutions to an optimal control problem, rather than a particular solution for a particular start and goal,” he said. “The mathematics we use leads to a very concise representation of the shape of the objects we’re trying to manipulate. The tools we use are very classical, but we’re using them in a completely different way.”
As a motivational case study, the group will be working to enable an automated installation of a wire harness, which is a bundle of wires that terminates in electrical connectors in a hybrid electric vehicle, by a commercially available manufacturing robot named Baxter. Baxter is a low-cost robot that is teachable by physical demonstration and is designed for use in small-to medium-sized businesses to automate handling and assembly of parts.
Baxter is already equipped to plug in the connectors, which are rigid objects, but the challenge will be manipulating the flexible wires. The team will be developing new algorithms for manipulation and perception of deformable objects, by modeling the wire harness as an elastic rod with variable stiffness whose centerline has a tree-like structure, called a branched elastic rod.
Eventually applications for this research include industrial manufacturing with robots, where the parts being assembled are flexible. Additionally, in the future, robots could be trained to manipulate flexible circuit boards or bundles of wires connecting circuit boards or possibly robotic surgery, where robots could be in charge of tasks such as automated suturing without human assistance.
“A lot of basic research needs to be done before robots can do these things, but we’re working on it,” Bretl said. “Often, the assembly of flexible parts is done by human workers and we’d like to be able to automate that.”