The Commission for the Conservation of Antarctic Marine Living Resources (CCAMLR) aims to develop a feedback approach to aid Ecosystem Based Management (EBM) of the Antarctic krill fishery in the Southern Ocean. Feedback approaches in fisheries management usually include harvest control rules (HCRs) which are derived from control theory. Model Predictive Control (MPC) optimises control rules based on their predicted performance in a model of the controlled system. Here we use MPC to develop a robust control strategy for a simulation model of a spatially resolved predator-prey system. We identify optimal HCRs based on an objective for the state of the target stock and constraints equivalent to limit reference points for both the target stock and its predators. Our results demonstrate that an approach based on optimisation is more likely than the current fixed catch limit to achieve CCAMLR’s EBM objectives. The MPC design illuminates the prerequisites for this sort of feedback management approach namely: clearly defined objectives and attitudes to risk; monitoring that includes the highest trophic level for which objectives are defined; and reliable models of system uncertainty. The approach also readily evaluates the trade-offs between objectives given the relevant levels of uncertainty. We conclude that MPC is a promising approach for addressing spatially-resolved multiple-objective problems with high levels of uncertainty, but that its information requirements demand both commitment to monitoring and clarity about objectives.