Visual Foresight of The Way to Robots' Independence

In a new development, researchers have introduced a new technology called "Visual Foresight." Visual foresight enables robots to predict the consequence of their future actions so as to manipulate 'strange' objects easily.

Researchers from UC Berkeley developed the learning technology that is about to change the world of artificial intelligence. Of course, it would also help self-driving cars to predict future events and also augment more intelligent robotic assistance. But this early prototype is just to teach them independent manual skills. Simple predictions that can only stretch up to several seconds but are still enough to make them compute how to successfully move objects on a table without getting in the way of obstacles. These tasks can be carried out autonomously, reason being that their visual imagination is learned unsupervised, from sratch. The robot plays with objects on the table after which it builds a predictive world model and use this information to operate new objects.

The research team performed a demonstration of this new technology yesterday at the Neural Information Processing Systems conference in Long Beach, California.Sergey Levine, an assistant professor in Berkeley's Department of Electrical Engineering and Computer Sciences said. "In the same way that we can imagine how our actions will move the objects in our environment, this method can enable a robot to visualize how different behaviours will affect the world around it"

"This can enable intelligent planning of highly flexible skills in complex real-world situations."Chelsea Finn, a doctoral student in Levine's lab- where the technology was developed-, who also invented the original DNA model also said "In the past, robots have learned skills with a human supervisor helping and providing feedback. What makes this work exciting is that the robots can learn a range of visual object manipulation skills entirely on their own.

"In a nutshell, this new learning technology is meant to augument artificial intelligence. They learn simple tasks and successfully interact with them. Even Frederik Ebert, a graduate student who also worked on the project said that Humans learn object manipulation skills without any teacher through millions of interactions with a variety of objects during their lifetime.

We have shown that it is possible to build a robotic system that also leverages large amounts of autonomously collected data to learn widely applicable manipulation skills, specifically object pushing skills."This is well under way and its future looks promising as more researches are currently ongoing.