Detailed processing could teach robots with manly tasks

For many people doing the household activities is a nightmare. It becomes an inseparable part of your life, and you feel pretty bored and helpless while doing such unproductive activities. Even if you do such household works, you do them with little importance. In such a situation how will you feel if a robot can help you with all your household activities?

In recent ongoing research, computer scientists have been making full efforts on educating the machines to perform the more significant amount of tasks around the house. As per the latest publications which are headed by the MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) along with the University of Toronto researchers. The researchers exhibited “Virtual Home” which is a system that can simulate in-depth household tasks and then leave the task to the artificial “agents” for correctly executing them.

In the first step, the team trained the system near about 3000 programs about various activities which are subsequently broken down into small subtasks to enable the machine to understand the same. An easy task like “making coffee” would also include the step of “grabbing a cup.” Keeping in pace with this statement, the researchers demonstrated a virtual home in a 3D world which is very much inspired by the Sims video game. It is learned that the team’s AI agent can go through 1000 of these interactions within the Sims-style world which involved eight different scenes starting from the living room, kitchen, dining room, bedroom and home office.

According to Ph.D. student Xavier Puig, in a computer program, the actions have to be described in an unambiguous way which will enable the machine to complete the task correctly. Xavier was the in charge in respect of this paper. He further added that such programs could train a robot or even a virtual character to be used as a representative for executing complex tasks with smooth actions.

This particular project was co-developed by CSAIL along with the University of Toronto and also the scientists from the McGill University. This latest project will be displayed at the Computer Vision and Pattern Recognition conference which will take place in the current month in the Salt Lake City.

Robots unlike the human beings need more detailed instructions to complete easy tasks. The robots cannot just interpret on their own quickly and so requires minute programming.