Humans have a way of understanding other people’s goals, desires, and beliefs, which is a crucial skill that allows us to anticipate people’s actions. Take the bread out of the toaster? You will need a plate. vacuuming papers I’ll take the green trash.
This skill, often referred to as “theory of mind,” comes easily to us as humans, but it’s still a challenge for robots. But if robots are to become truly helpful assistants in manufacturing and in everyday life, they need to learn the same capabilities.
In a new paper, computer science researchers at USC Viterbi, finalists for Best Paper at the ACM/IEEE International Conference on Human-Robot Interaction (HRI), aim to teach robots how to predict human preferences in assembly tasks, so that they can One day of help with everything from building a satellite to setting a table.
“When working with people, the bot needs to constantly guess what the person is going to do next,” said lead author Hirambe Nimelikar, a computer science PhD student at USC working under Stefanos Nikolaidis, assistant professor of computer science. “For example, if a robot thinks a person will need a screwdriver for the next part assembly, it can get the screwdriver in advance so the person doesn’t have to wait. This way the robot can help people finish the assembly faster.”
But, as anyone who has shared furniture with a partner can attest, predicting what a person will do next is tricky: Different people prefer to build the same product in different ways. While some people want to start with the hardest parts to get over it, others may want to start with the easiest parts to save energy.
pridect
Most current technology requires people to show the robot how they would like the assembly to perform, Nemlekar said, but that takes time and effort and can defeat the purpose. “Imagine you have to put together a whole plane just to teach the robot your preferences,” he said.
But in this new study, the researchers found similarities in how an individual aggregates the different products. For example, if you started with the hardest part when building the Ikea sofa, you’ll likely use the same tact when assembling the crib.
So, instead of “showing” the bot its preferences on a complex task, they created a small collection task (called a “basic” task) that people could easily and quickly perform. In this case, parts of a simple model of the aircraft are assembled, such as wings, tail and propeller.
The robot “watched” the human completing the task using a camera positioned directly above the assembly area, looking down. To detect human-operated parts, the system used April tags, similar to QR codes, attached to the parts.
Then, the system used machine learning to learn a person’s preference based on the sequence of actions in the primary task.
“Based on how the person does the small assembly, the robot predicts what that person will do in the larger assembly,” Nemlikar said. “For example, if a bot sees that someone likes to start the small assembly with the easiest part, it will expect that it will also start with the easiest part of the large assembly.”
Building confidence
In the researchers’ user study, their system was able to predict the actions humans would take with up to 82% accuracy.
“We hope our research will make it easier for people to show their preferences to robots,” said Nemlikar. “By helping each person in their preferred way, bots can reduce their work, save time, and even build trust with them.”
For example, imagine that you are assembling a piece of furniture at home, but you are not particularly helpful and are struggling with the task. A robot that has been trained to predict your preferences can provide you with the necessary tools and parts ahead of time, making assembly easier.
This technology could also be useful in industrial settings where workers are tasked with assembling products on a large scale, saving time and reducing the risk of injury or accidents. In addition, it can help people with disabilities or limited mobility assemble products more easily and maintain independence.
Preferences quickly learn
The goal, researchers say, is not to replace humans on the factory floor. Instead, they hope that this research will lead to significant improvements in the safety and productivity of assembly workers in human-robot hybrid factories. “Robots can perform non-value-adding tasks or challenging tasks that are currently being performed by workers.
For the next steps, the researchers plan to develop a method to automatically design core tasks for different types of collection tasks. also aimed to assess the usefulness of learning human preferences from short tasks and predicting their actions on a complex task in different contexts, for example, personal assistance in the home.
“While we’ve observed human preferences move from basic to actual tasks in assembly manufacturing, I expect similar results in other applications as well,” Nikolaidis said. “A robot that can quickly learn our preferences can help us prepare a meal, rearrange furniture, or make home repairs, making a huge impact in our daily lives.”