Agglomeration, which leads to growth in particle size over time, is an important phenomenon in several applications that use fluidized bed technologies. Understanding, predicting and controlling this phenomenon are important to avoid operational problems, reduced efficiencies, reactor down-time and thereby monetary losses. Operating conditions, material properties and particle hydrodynamics can simultaneously affect the extent of agglomeration. This makes accurate prediction with mere experimentation difficult. Theoretical models to predict agglomeration based on particle collision dynamics have been evolving. They will enable easier predictions with quick estimates for any changes in parameters. In general, the models use certain criteria to determine the outcome of two-particle collisions, and a technique to estimate the kinetics of agglomerate growth. In addition, they may also include a method to account for the hydrodynamics of the discrete and continuous phases. This paper provides a critical review of the development of such theoretical models and analyzes their suitability to different applications. Several avenues for further development of mathematical models to predict agglomeration are identified. Published by Elsevier B.V.