Communications of the ACM
MIT Engineers Devise Recipe for Improving Any Autonomous Robotic System
Part of a hardware demonstration in which the tool automatically optimizes the performance of two robots working together to move a heavy box.
Credit: Charles Dawson et al.
A general-purpose optimizer developed by Massachusetts Institute of Technology (MIT) researchers aims to speed up the development of autonomous robotic systems by applying an optimization code to simulations that automatically determines where and how to tweak the system to boost performance.
Said MIT's Charles Dawson, "Instead of saying, 'Given a design, what's the performance?' we wanted to invert this to say, 'Given the performance we want to see, what is the design that gets us there?'"
The researchers tested the tool, which is based on automatic differentiation, on a system in which a wheeled robot plans paths between two obstacles using beacons.
The tool identified the best placement of the beacons within five minutes, versus 15 minutes using conventional methods.
It also completed the optimization process 20 times faster than conventional methods when applied to a system in which two wheeled robots collaborated to move a box.
From MIT News
View Full Article
Abstracts Copyright © 2022 SmithBucklin, Washington, DC, USA
No entries found