Tech Peek | Self Learning Robots
2:01
Fried chicken chunks might not be the first thing that comes to mind after you hear the phrase “deep reinforcement learning,” but asking a robot to build a Japanese bento box meal is exactly one of the tasks being used to develop self-learning robots. Why are chicken parts challenging? Because no two look the same. To solve the problem, San Francisco-based Osaro developed software that integrates vision systems, robots and deep reinforcement learning. International fast-food cuisine provides a structured learning environment, as does picking random, loose objects in a bin as part of an automated storage and retrieval systems. By learning in these mundane applications, eventually robots will be able to autonomously handle more challenging and complex manufacturing tasks like welding or moving hot castings. But for now, pass the soy sauce.
