Robotics firm Agility Robotics has enabled its bipedal walking robot Cassie to take on stairs and stair-like structures.

Using proprioception, which is the perception of limbs and how they are used in environments without vision — think touching one’s nose with one’s finger while their eyes are closed — the team enabled Cassie to mount and descend stairs. According to Agility Robotics, this approach uses just a fraction of the processing power that would be necessary to accomplish such a task using lidar sensors commonly employed in autonomous vehicles or other robots.

Source: Agility Robotics

Cassie was taught to climb and descend stairs via reinforcement learning wherein the robot was trained on a dataset of staircase scenarios. Additionally, Cassie is also capable of maneuvering over unstructured terrain and up and down steep hills.

Agility Robotics envisions future applications of the bipedal stair-climbing robot could include parcel delivery or elder care applications. The robot is described in the journal Computer Science.

To see how the bipedal robot mounts and descends stairs, watch the accompanying video that appears courtesy of Agility Robotics.