Researchers at Utsunomiya University in Japan have written a mathematical model for the sport of wrestling, which they have tested in a numerical simulation.
The model defines wrestling as a system of two mechanical agents connected through actions such as contact and collision, with the objective being for one agent to knock over the other while maintaining its balance.
The robot wrestlers are depicted as inverted pendulums on a cart that can move backwards or forwards, linked to each other by a spring that can stretch and retract.
The researchers examine how best to design an intelligent controller that outperforms its opponent, with its action limited to moving the cart backwards or forwards.
The team found 17 parameters that influence wrestler behavior, such as pendulum mass and length, cart mass, and acceleration due to gravity. The controller must determine how to move to maintain the upright position of its own pendulum while exerting a turning force to unbalance the other player. Given the researchers' parameters, the controllers were unable to simulate solutions successfully and winners were essentially selected at random. The team resolved this by adding a short delay to calculations, which prevents controllers from attaining levels of complexity that interfere with solutions.
In the future, the researchers would like for the controllers to compete in a human versus machine contest.
From Technology Review
View Full Article
Abstracts Copyright © 2014 Information Inc., Bethesda, Maryland, USA
No entries found