Cassie Robot Runs a 5K - Bipedal Robotics Achievement

Cassie: From Ostrich-Inspired Robot to 5K Runner

Many will recognize Cassie as the foundational technology behind Agility Robotics’ delivery robot, Digit. Those closely following the advancements in robotics are likely aware that Cassie initially began as a design inspired by the ostrich, consisting primarily of legs.

Origins and Development

Beyond its role in propelling Agility’s commercial goals, the robot has served as a robust platform for investigating bipedal locomotion. Cassie was originally conceived by Jonathan Hunt, a professor at Oregon State University, in 2017.

The project was initiated with a $1 million grant from the Department of Defense – a common scenario within the robotics field.

Achieving a 5K Run

Recently, researchers from the Dynamics Robotics Laboratory at OSU’s College of Engineering showcased their ongoing efforts to maximize Cassie’s bipedal capabilities.

The team successfully demonstrated Cassie completing a 5K run completely untethered, powered by a single battery charge. While not challenging human athletic records, the 53-minute and three-second run represents a significant achievement for the technology.

During the run, approximately 6.5 minutes were dedicated to resolving technical issues, including an overheated computer and a miscalculated turn that caused the robot to fall.

Self-Learning and Efficiency

“Cassie’s design and construction contribute to its efficiency, allowing us to truly test the boundaries of its hardware and demonstrate its potential,” stated Jeremy Dao, a Ph.D. student involved in the project.

The robot effectively learned to run through the application of a deep reinforcement learning algorithm. This allowed Cassie to autonomously determine how to maintain stability while in motion by dynamically adjusting its balance.

“Deep reinforcement learning is a potent AI technique that enables skills such as running, skipping, and navigating stairs,” explained Yesh Godse, an undergraduate student on the team.

Further Advancements

In May of the same year, the OSU team also exhibited Cassie’s ability to ascend and descend stairs without relying on lidar or onboard cameras.

This demonstrates the robot’s increasing autonomy and adaptability in complex environments.