Optimizing the heating ventilation and air condition (HVAC) control process generally considers a trade-off between thermal discomfort and energy cost. Technically, this can be represented as a multi-objective optimization problem, where the occupants’ thermal preferences and the cost of energy consumed are two conflicting objectives. When the occupancy schedule is known in advance this problem typically reduces to minimizing the energy cost while meeting the thermal comfort constraint set by a user, typically defined as a setpoint-temperature-based comfort band. However, this approach is inadequate in dynamic occupancy settings where the optimization process is driven by occupancy estimates that are inherently uncertain. In particular, ensuring that the room temperature lays within a typically narrow comfort band, even when there is small probability of occupancy, significantly increases HVAC energy usage. In this case, accepting some probability of discomfort, is essential to reduce energy cost. We investigate three optimization HVAC control algorithms that deal with dynamic occupancy and we provide a comparative analysis to show the advantages and disadvantages of each approach with respect to their efficiency, effectiveness, applicability and usability in different settings.