FCMAC based on minesweeping strategy
Jingsong Hu, Guiwu Hu, Jiabing Wang, Bo Liu · 2005
The fuzzy cerebellar model articulation controller has been widely used to control complex object, but most learning algorithm of CMAC are greedy and local optimization. So the precision is relative low. A fast global strategy, call minesweeping strategy, are presented to improve the global ability of CMAC. The minesweeping strategy lets the current search "jump out", rather than "climb out" step by step, hardly and wanderingly, as simulating annealing and tabu search do, from the current local minimum by exploiting a new area that is far away from all obtained local minima erenow. Therefore the strategy to solve local minimum problem is more successful and faster than other methods. The new method realizes fine control quality to a nonlinear plant, of which mathematic model is not known.