Table 5.

Control algorithms for drug administration.

Control Objective	Algorithm	Advantages	Disadvantages	Applications
Maintain drug concentration at certain levels	PID control	Easy to implement; flexible structure for functional expansion.	Design of PID controllers usually needs linear models. The linear approximation of the nonlinear model leads to information loss that eventually decreases control performance.	For anesthesia delivery146. For designing anti-retroviral therapy for HIV165.
Balance between drug toxicity, drug cost and therapeutic performance	Optimal control	Balance drug cost, toxicity and therapeutic performance.	More computations are required. Weighting variables Q and R are sensitive and they should be adjusted carefully.	For optimal imatinib treatment for leukemia model174. For eliminating cancer cells using abiraterone163.
Handle constraints in the model; Balance between drug toxicity, drug cost and therapeutic performance	Model predictive control	Consider physical constraints based on drug toxicity and patient conditions.	Solving the constrained optimization problem needs additional work to ensure stability, optimality and feasibility.	For stabilizing HIV infection165. For anesthesia delivery with uncertainty175.