Jump to content

Robots Learn How to Work Together More Efficiently


Recommended Posts

For humans, when something is too difficult to do we can enlist the help of others to get the job done, but for robots it is not as simple. For multiple robots to work together their actions may need to be coordinated and the results can be very efficient, in a controlled environment. In other settings, the efficiency can vanish because of how complex the computational work is, but researchers at MIT have found a solution to this problem.

Instead of requiring the robots to build a comprehensive plan, this new method breaks it up into single actions. At first these actions are evaluated to determine if the one robot involved will be able to complete the task successfully, and then checked to see if they are compatible with the steps before and after, and the additional robots involved in those steps. If the robots cannot determine a viable solution to the step, it will just skip the step and move on, coming back to it later. This deferment is actually critical to the method as it allows the actions to be interrupted at any point, and it also allows the parts the robots are carrying to be freely dropped and regrasped as needed.

The resulting sequence of actions may not be the optimal solution, but it works and can be found significantly faster. In fact some optimal solutions would take thousands of years of computation to discover, so by just finding a viable solution, this method should prove very powerful for robots in the real-world, outside of factories and such.

 

 

Source: MIT



Back to original news post

Share this post


Link to post
Share on other sites

×
×
  • Create New...