When it comes to the evolution of mobile robots, it may be a long time before two-legged robots can interact safely in the real world, according to a new study.
Led by a team of researchers at Ohio State University, the study was recently published at the IEEE/RSJ International Conference on Robotics and Intelligent Systems (IROS) 2022, and describes a framework for testing and characterizing the safety of two-legged robots, machines that differ from their own. Their wheeled counterparts, rely on mechanical limbs for locomotion. The study found that many existing robotic models do not always act predictably in response to real-life situations, which means that it is difficult to predict whether they will fail — or succeed — at any given task that requires movement.
“Our work reveals that these robotic systems are complex and, most importantly, counterintuitive,” said Bowen Wing, an electrical and computer engineering doctoral student at Ohio State. “This means that you can’t rely on the robot’s ability to know how to act in certain situations, so the completion of the test becomes more important.”
With the development of mobile robots to carry out more diverse and complex tasks, many in the scientific community have also noted that the industry needs a set of global safety testing regulations, especially as robots and other artificial intelligence are gradually beginning to flow into our daily lives. Wong said that legged robots in particular, which are often made of metal and can run at speeds of up to 20 miles per hour, can quickly become safety hazards when they are expected to work alongside humans in real and unexpected environments. many times.
“Testing is really about assessing risk, and our goal is to investigate how much risk the bots currently pose to users or customers in action,” he said.
While there are currently some safety specifications in place for deploying two-legged robots, Weng noted that there is not yet any common agreement on how to test them in the field.
This study develops the first data-driven, scenario-based safety-testing framework of its kind for two-legged robots, Wong said.
“In the future, these robots may have the opportunity to live with humans side by side, and they are likely to be produced collaboratively by several international parties,” he said. “So having safety and testing regulations in place is critical to the success of this type of product.”
The research, inspired in part by Weng’s work as a vehicle safety researcher at the Transportation Research Center, which partners with the National Highway Traffic Safety Administration, leverages sample-based machine learning algorithms to see how robot simulations will fail while testing the world.
Although many factors can be used to describe a robot’s overall safety performance, this study analyzed a range of conditions under which a robot would not fall while actively navigating a novel environment. Because many of the algorithms the team used were derived from previous robotics experiments, they were able to design multiple scenarios to run the simulations.
One experiment focused on examining the robot’s ability to move while performing tasks in different gaits, such as walking backwards or walking in place. In another study, researchers tested whether a robot would tumble if it was pushed periodically with enough force to change its direction.
The study showed that while one robot failed to stay upright for 3 out of 10 trials when asked to speed up its gait slightly, the other could stay upright over 100 trials when pushed from its left side, but fell during 5 out of 10 trials when the same force was applied. on her right side.
Ultimately, the framework could help the researchers certify commercial deployment of two-legged robots and help establish a safety standard for robots built with different structures and characteristics, though Weng noted that it will take some time before it can be implemented.
“We believe that this data-driven approach will help create an unbiased and more efficient way to conduct observations of bots in test environment conditions,” he said. “What we are working on is not immediate, but for future researchers.”
Co-authors are Guillermo Castillo and Ayonga Hereid from Ohio State and Wei Zhang from the Southern University of Science and Technology in Shenzhen, China. This work was supported by the National Science Foundation and the National Natural Science Foundation of China.