A variable structure model (VSM) was used for online control of a bench-scale anaerobic digestor. The VSM consisted of three submodels, which described the methanogenic, organic overload, and acidogenic states of the digestion process. Submodel activation was handled by a knowledge-based system, which used online measurements of the digestor pH and biogas composition to identify the process state and thereby select an appropriate leading submodel. Separate controllers were designed for each submodel, but only the controller corresponding to the leading submodel was applied at each instant. Experimental validation of the VSM-based feedback process control in a 5-L anaerobic digestor showed a stable process performance under various operating conditions.