Reduced-order model predictive control of a fish schooling model

We study the problem of model predictive control (MPC) for fish schooling proposed by Gautrais et al. (2008). The high nonlinearity attributed to its attraction/alignment/repulsion law suggests need use MPC controlling schooling’s motion. However, large schools, hybrid nature can make it numerically demanding perform finite-horizon optimizations in MPC. Therefore, this paper proposes reducing efficient MPC; reduction is based on using weighted average directions individual school. analytically show how normalized eigenvector centrality alignment-interaction network yield a better comparing errors. confirm finding weight and numerical efficiency with reduced-order simulations. allows us school up 500 individuals. Further, we that improve accuracy factor five when compared constant weights.