Event-triggered drug dosage control strategy of immune systems via safe integral reinforcement learning

Abstract

This paper develops an event-triggered drug dosage control strategy for immune systems with state and input constraints via safe integral reinforcement learning. By developing a novel performance index function with control barrier function included, the state and control input can be constrained within specified ranges, that is, the number of cells and the dosage of the drug can be maintained within the appropriate range, which ensures the safety of the immune system. Subsequently, the drug dosage control strategy is designed under the event-triggered mechanism, which means that it is updated only when necessary instead of periodically adjusted over time, effectively reducing the number of drug dosage control strategy adjustment and saving computational resources. Moreover, a single critic neural network structure is established to attain an approximate event-triggered drug dosage control strategy. Theoretical analysis shows that under the developed novel event-triggered condition, the approximate event-triggered drug dosage control strategy ensures the tracking error is uniformly ultimately bounded. Ultimately, the simulation experiments confirm the effectiveness of the proposed safe integral reinforcement learning based event-triggered drug dosage control strategy. Note that this drug dosage control strategy can adjust the quantities of pathogens and immune cells to the desired levels with minimal drug costs, achieving precise treatment while reducing drug side effects.