How do you improve an automated planner such as a planning algorithm? By adding human intuition into the mix, of course – that is according to researchers from MIT’s Computer Science and Artificial Intelligence Laboratory.
Planning algorithms are amazing – they can do anything from scheduling flights to coordinating tasks for teams of autonomous satellites. However, even the best of them aren’t as effective as humans with a talent for problem-solving. For this reason, MIT researchers are trying to improve planning algorithms by giving them the benefit of human intuition.
To add human intuition to automated planners, the researchers encoded the strategies of 36 high-performing human planners in a machine-readable form. Then, they fed the algorithms these strategies, improving the performance of competition-winning planning algorithms by 10% to 15% on a challenging set of problems.
“The plan that the planner came up with looked more like the human-generated plan when it used these high-level strategies from the person,” said Julie Shah, an assistant professor of aeronautics and astronautics at MIT.
Essentially, the machine took the insights and high-level strategies of people who are excellent at planning thereby improving its own problem-solving.
The team is now working on making the system fully automatic by using natural-language processing techniques. The goal is to create an algorithm that will be able to convert users’ free-form descriptions of their high-level planning strategies into linear temporal logic – all on its own.
MIT – Massachusetts Institute of Technology (http://news.mit.edu/2017/human-intuition-planning-algorithms-0207)