To explore and utilize the advantages of droplet-based microfluidics, hydrodynamics, and mixing process within droplets traveling though the T junction channel and convergent-divergent sinusoidal microchannels are studied by numerical simulations and experiments, respectively. In the T junction channel, the mixing efficiency is significantly influenced by the twirling effect, which controls the initial distributions of the mixture during the droplet formation stage. Therefore, the internal recirculating flow can create a convection mechanism, thus improving mixing. The twirling effect is noticeably influenced by the velocityof thecontinuous phase; in the sinusoidal channel, the Dean vortices and droplet deformation are induced by centrifugal force and alternative velocity gradient, thus enhancing the mixing efficiency. The best mixing occurred when the droplet size is comparable with the channel width. Finally, we propose a unique optimized structure, which includes a T junction inlet joined to a sinusoidal channel. In this structure, the mixing of fluids in the droplets follows two routes: One is the twirling effect and symmetric recirculation flow in the straight channel. The other is the asymmetric recirculation and droplet deformation in the winding and variable cross-section. Among the three structures, the optimized structure has the best mixing efficiency at the shortest mixing time (0.25 ms). The combination of the twirling effect, variable cross-section effect, and Dean vortices greatly intensifies the chaotic flow. This study provides the insight of the mixing process and may benefit the design and operations of droplet-based microfluidics.