Bowen Weng: Making Legged Robots Safe

April 17, 2024

Think about when someone is going on a hike. Humans have an innate ability to adapt to the ever-changing environment we encounter along the trail. We navigate rougher terrain, adjust to steeper inclines, appropriately respond to unpredictable wildlife, and circumvent fallen tree branches without too much thought.

However, the hike would not be quite so effortless for a robot. Robots use programmed algorithms and mechanical systems to perceive and respond to their surroundings. Reacting to those things for robots is complex and goes beyond a mathematical formula.

Besides getting the robot to go on the hike, one of the goals for the robot is to be able to be integrated alongside other humans. Now, imagine a robot with legs going on the hike alongside yourself. How can you ensure that it’s safe for you to interact with?

In an ideal world, the robot would have undergone tests – like we do with cars or any other type of machinery. However, the unique operating mechanisms of legged robots mean that it is difficult to create standardized testing methods. Different-legged robots operate in a variety of manners, leading to differences in their structures and movement patterns. Therefore, there’s no one way to test them and that is why ensuring their safety is a challenge - one that researchers such as Bowen Weng, Assistant Professor in the Department of Computer Science at Iowa State University, are working on solving.

“To the best of our knowledge, legged robots have not led to any known fatalities across the world,” Weng said while discussing the long-term implications of his research. “We would like to keep it that way, and even if this is statistically inevitable, we would like to do as much as possible to delay the occurrence of that moment.”

So, how does Weng go about trying to ensure that the robots he’s working on are safe for human interaction?

“I work on two (contradictive) safety perspectives safety perspectives of legged robots,” Weng added when explaining his research. “On the one hand, I build safe robots from a developer’s perspective. On the other hand, I also build testing frameworks to check and ensure the developer’s claim from a third-party point of view.”

Revolutionizing Safety Testing

Introducing an innovative approach to safety testing for legged robots, his paper  “On Safety Testing, Validation, and Characterization with Scenario-Sampling: A Case Study of Legged Robots”  was accepted to the IEEE/RSJ International Conference on Intelligent Robots and Systems 2022 (IROS 2022). IROS is one of the largest and most prestigious conferences in robotics, providing a platform for researchers, engineers, and practitioners to present their latest research findings in robotics and automation.

In the paper, they suggested a new way to test a robot’s safety. They proposed a method that involves testing the robot in a few different scenarios – allowing us to understand specific conditions where the robot is safe and how safe it is in those conditions. Such a framework can ensure that a robot is safe for use in the real world alongside humans and allow researchers to compare the safety performance of different-legged robots with different structures and ways of moving.

To test their proposed method, Weng and his team applied it to various advanced-legged robots – such as those that use dynamic models, deep neural networks, and reinforcement learning. They also accounted for different terrains, such as moving on slopes (like up and down hills), sudden changes in speed, and handling unexpected disturbances. Their results show that their method is an effective way of assessing the overall safety and performance of the robots.

Continued Research and Refinement

Weng and his team continued the research on testing algorithms for general robots with the IEEE Transactions on Robotics paper “On the Comparability and Optimal Aggressiveness of the Adversarial Scenario-Based Safety Testing of Robots.” They challenged the idea that prioritizing high-risk scenarios improves testing efficiency.

“We essentially showed there is a “No-Free-Lunch” theorem in safety testing. The NFL theorem is a classic concept and topic in searching algorithms and optimization,” explained Weng.

Further research was conducted, as presented in the paper “Towards Standardized Disturbance Rejection Testing of Legged Robot Locomotion with Linear Impactor: A Preliminary Study, Observations, and Implications,” which was accepted to the IEEE International Conference on Robotics and Automation (ICRA) 2024, the flagship conference in robotics.  In the paper, Weng uses a linear impactor to create disturbances that test how well a legged robot can handle unexpected situations. The linear impactor works to standardize testing as it is equitable to all-legged robots, can be repeated, and is adaptable. Their results showed that using the linear impactor is a promising way to evaluate and compare the ability of legged robots to handle disturbances.

The research was conducted in collaboration with Guillermo Castillo, a Ph.D. student from the Department of Electrical and Computer Engineering at The Ohio State University, Yun-Seok Kang, an Associate Professor from the Division of Health Sciences in the School of Health and Rehabilitation Sciences at The Ohio State University, and Ayonga Hereid, Assistant Profession in the Department of Mechanical and Aerospace Engineering at The Ohio State University.

Leadership in Legged Robot Safety Standards

Weng’s research on testing legged robot safety and establishing safety is highly regarded. Outside of his research, he is working on defining the future of safety standards with ASTM International. This international standards organization develops and publishes technical standards for various materials, products, systems, and services. Weng chairs ASTM Subcommittee F45.06 Legged Robot Systems.

The new subcommittee works on developing standards and testing procedures to evaluate the performance of mobile robots that primarily rely on articulated limbs and legged mechanisms to move around.  As a result, the standard will provide testing methods and performance metrics. These metrics can then be used to help lay the groundwork for regulatory frameworks that will uphold high standards of safety and accountability – an important first step in making sure that our future is safe.

Weng’s commitment to ensuring the safety and well-being of individuals in an increasingly automated world will be crucial for shaping a safe and sustainable future where humans and robots can live in harmony. With his work, we can engage with legged robots, whether on a hiking trail or in the workplace, knowing they are safe for us to interact with.

To learn more about his research, please watch the video at the top.

News Type: