Model predictive algorithms based on fuzzy discrete alternatives

Abstract

The application of model predictive algorithms to nonlinear processes results in a non-convex optimization problem for computing the optimal inputs. The optimization problem can be solved by using discrete search techniques, such as branch-and-bound, which has been applied to predictive control. However, for computational reasons a small number of possible discrete inputs must be used, which results in poor control performance. A possible solution to this problem is the use of adaptive input alternatives based on fuzzy rules, which were proposed previously to solve this problem in predictive control problems. The paper generalizes fuzzy discrete alternatives to predictive algorithms. An example of a robot were the position references are derived using the fuzzy-rule based optimization is presented, revealing good control performance.