Sorry, came late to the party, but I'll drop my question anyway: Don't you think the old fashion theorem/model making is slowly getting replaced by exhaustive simulation/search based approaches, enabled by our ever faster computers that can handle gazillions of data? Take the famous cart-pole balancing problem for instance, sure we can tune some parameters and let the computer find the answer, instead of going through the trouble of modeling friction, performing linearizations, calculating a control law and so on. But in this case, aren't we missing the point sometimes? As a leading researcher on the field of evolutionary robotics, I would love to hear your thoughts on this issue.
tioguerra2 karma
Sorry, came late to the party, but I'll drop my question anyway: Don't you think the old fashion theorem/model making is slowly getting replaced by exhaustive simulation/search based approaches, enabled by our ever faster computers that can handle gazillions of data? Take the famous cart-pole balancing problem for instance, sure we can tune some parameters and let the computer find the answer, instead of going through the trouble of modeling friction, performing linearizations, calculating a control law and so on. But in this case, aren't we missing the point sometimes? As a leading researcher on the field of evolutionary robotics, I would love to hear your thoughts on this issue.
View HistoryShare Link