In IoT Sensor Networks, majority of MAC protocols did not consider the quality of wireless links. Bidirectional and concurrent transmissions occur more often than traditional WSN. The sleep/wakeup period should adaptively varied based on the variance in the network loads. In this paper, we propose a Multi-population Firefly Algorithm (MFA) based MAC protocol for Dynamic Sleep Scheduling in Clustered IoT Sensor Networks. In the system model, the IoT devices are grouped into clusters based on their requirements and architecture. The cluster heads are randomly selected in each cluster based on their remaining energy levels. The sleep wakeup period of each node adaptively fixed using Multi-population Firefly Algorithm (MFA) based on the Energy Level, Expected Load and Channel quality parameters. Simulation results show that the MFA MAC protocol has lesser energy consumption with higher packet delivery ratio, when compared to existing work.