4 ms·
The answer, with a diagram, can be found here: http://mathworld.wolfram.com/SpiderandFlyProblem.html http://mathworld.wolfram.com/SpiderandFlyProblem.html
by mayoff 15y ago
The answer, with a diagram, can be found here:
http://mathworld.wolfram.com/SpiderandFlyProblem.html http://mathworld.wolfram.com/SpiderandFlyProblem.html
- losvedir 15y agoHa, well that was a lot easier than I made it. I ended up with the right answer by finding the minimum of the equation f(x,y) = 2(sqrt(1+x^2) + sqrt((6-x)^2 + y^2) + sqrt((15-y)^2 + 36)) which is the length of the path taken if the ant goes up to the ceiling at a point x to the East of straight up, and from there goes to the side wall at a point y to the south, and then goes from there to the center of that wall. (And then it's mirrored). I should have thought to unravel the room! Errr.
- d0mine 15y agofor the unraveled room: sqrt(1+x^2) + sqrt((6-x)^2+y^2)+sqrt(12^2+z^2)+ sqrt((30-z-y)^2+u^2)+sqrt((6-u)^2+1) 40 at {x -> 0.75, y -> 7, z -> 16, u -> 5.25} http://www.wolframalpha.com/input/?i=minimize%5B%7Bsqrt%281%2Bx%5E2%29+%2B+sqrt%28%286-x%29%5E2%2By%5E2%29%2Bsqrt%2812%5E2%2Bz%5E2%29%2Bsqrt%28%2830-z-y%29%5E2%2Bu%5E2%29%2Bsqrt%28%286-u%29%5E2%2B1%29%2Cz%3E%3D0%2Cy%2Bz%3C%3D30%2C0%3C%3Dx%3C%3D6%2C0%3C%3Dy%3C%3D30%2Cu%3E%3D0%2Cu%3C%3D6%7D%2C%7Bx%2Cy%2Cz%2Cu%7D%5D http://www.wolframalpha.com/input/?i=minimize%5B%7Bsqrt%281%... your equation: http://www.wolframalpha.com/input/?i=minimize+2%28sqrt%281%2Bx%5E2%29+%2B+sqrt%28%286-x%29%5E2+%2B+y%5E2%29+%2B+sqrt%28%2815-y%29%5E2+%2B+36%29%29 http://www.wolframalpha.com/input/?i=minimize+2%28sqrt%281%2...
- tzs 15y agoTo make the "minimize the equation" approach complete, other paths should be considered, such as instead of going up to the ceiling, going to the side and taking a side wall. Some limits can be placed on the number of possible paths by considering some general properties of an optimal solution: 1. Assuming the start point is A, and the path leaves the start side at point B, the optimal path will take a straight line from A to B. Proof: if it were not straight, it could be replaced by a straight line which would shorten the path, contradicting the assumption the path was optimum. 2. Once the optimal path leaves a side, it will not return to that side. Proof. Suppose the path leaves the side at point B, then later enters at C, and then leaves again at D. The route from B to C could be replaced with a straight line from C to D, which would shorten the path, again contradicting the assumption that the path was optimum. This cuts down the number of possible path templates greatly, since a given surface can only appear once in the optimal path, and only contain a single straight line segment. I expect that a little more reasoning can deduce some more limits on possible optimal paths, to get it down to just two or three possible equations to maximize.