Past literature (Inoue et al., 1993) has indicated that a significant amount of NOx can be reduced by operating a spark ignition (SI) engine at the stoichiometric condition with high exhaust gas recirculation (EGR) levels. However, the problem has been the engine instability due to cyclic dispersion in heat release (Sutton and Drallmeier, 2000). A novel neural network (NN) controller is developed to reduce the cyclic dispersion in heat release in SI engines with high levels of EGR when the engine dynamics are unknown. The NN controller can also be used to minimize engine emissions at extreme lean condition, where similar complex cyclic dynamics are observed.