The researchers from California Berkeley University developed a robot named Cassie, which can teach itself to walk. The robot has a pair of legs connected to a holding frame that operates the body. Its ability may seem simple but how they can teach itself is the new kind of advanced technology.
The technology used with Cassie is Reinforcement Learning. It allows the robot to learn from trial and error. The researchers first create the stimulation of the robot in a computer to train them without damaging the hardware. Cassie was set to achieve the goals to, for example, walk upright. It learned from achieving several goals and applying the lessons learned with one another. Once it learned to walk, the team transferred the information to the hardware and Cassie continued learning more. It learned from falling and slipping and how to avoid them. It even learned how to compensate when the motors are damaged.