Complex and drastic variations of atmospheric aerosol components lead to high uncertainties in climate change assessment. The remote sensing of aerosol composition is the technology often used in aerosol optical and microphysical parametric analysis. These parameters, which are derived from remote sensing measurements, can quantitatively estimate the aerosol components of the entire atmosphere. Remote sensing offers the advantages of real-time and fast detection, spatial coverage, and maintenance of natural aerosol status. This paper presents a comprehensive review of theories, observations, models, and algorithms for aerosol composition remote sensing. First, in the area of algorithm development, we analyze the main ideas and methods for establishing aerosol composition models. In particular, an advanced remote sensing categorization model, which includes black carbon, brown carbon, mineral dust, light-scattering organic matter, inorganic salts, sea salt, and water, is presented in detail. The methods are identified or distinguished according to components (i.e., sensitivity parameters, including light absorption, size distribution, particulate shape, etc.). Second, for the calculation of the refractive index, some of the typical methods suitable for different aerosol mixing states are compared, and then their impacts on composition retrieval are determined. Third, some examples of remotely sensed aerosol components are given, and a preliminary validation of the retrievals is conducted by using synchronized chemical measurements. Finally, the development tendencies of the remote sensing of atmospheric aerosol composition are summarized from the perspectives of observation capability enhancement, optimization of the categorization model, improvements in retrieval accuracy, extension of application abilities, and identification of utilization prospects in global climate change assessment.