This paper presents the use of control engineering principles to optimize mobile and wireless health (mHealth) adaptive behavioral interventions for physical activity based on Social Cognitive Theory (SCT). SCT is a conceptual framework that describes human behavior and has been used in many health behavior interventions. An intervention for physical activity is formulated as a control systems problem relying on a dynamical model of SCT that is developed utilizing fluid analogies. To obtain values for model parameters, system identification experiments are designed including two phases: An initial informative stage followed by an optimized stage that incorporates 'patient-friendly' conditions. With the estimated model, a closed-loop intervention is formulated relying on Hybrid Model Predictive Control (HMPC). The HMPC algorithm includes a representation of categorical and discrete constraints that are inherent to behavioral interventions, and the recognition of behavioral initiation and maintenance phases. A simulation study is performed illustrating representative scenarios of the system (in both open and closed-loop).