This robot doesn’t need direct programming for learning how to walk

Apr 16, 2021 | Vanshika Kaushik

This robot doesn’t need direct programming for learning how to walk title banner

Cassie is a two legged robot which doesn’t need direct programming to assist it in walking, it is developed by a team of researchers based in University of California.


This two legged robot holds the ability to teach itself how to walk.This has been made possible by Reinforcement Learning.


Cassie can lower its walking height, recover to normal height and can push return from back to front or front to back on its own.


No specific coding was done by programmers to teach this robot how to walk.This is a breakthrough invention.


Cassie appears to be a simpler model compared to the one invented by Boston Dynamics. It is  still a long way to witness the evolution of an entirely self-programmed  robot but this invention has made the far-sighted dream a little closer to reality.


Reinforcement Learning


Reinforcement Learning is a type of Supervised machine learning that teaches the agent to communicate with the environment.Robots require specialized programming for walking,bending up and bending down.


RL’s idea is very similar to how the human psychology for example a toddler when learns how to walk observe people around them.First they process basic walking later step by step toddlers learn how to stand and how to sit by observing people around them


The same is true for Cassie,she  was first taught to simulate walking in the Virtual World.The robot was trained not programmed .It was trained with information related to walking.The AI could remember the process of walking.


(Recommended read: Machine Learning Algorithms)


Further in the real world when the robot was tested it could walk on its own.It was due to an AI engine that  recollected  the process of Virtual World when it was placed in the real one.


According to Tech Explore Cassie was first given instructions on how to walk but Cassie didn’t stop there.She continued to learn. Within the passage of time Cassie even learnt how to work without her mortar. This was possible through deep AI as it enables the robots to learn by observing. 


Initiating the journey of progression by walking on her own Cassie can now lower its walking height and can even recover back to its normal height. 

Tags #Artificial intelligence#Machine learning