Physics-based simulation (PBS) is now widely utilized to maximize the usage of real-time sensory information in surveillance applications. Since agent-based simulation (ABS) helps in analyzing human behaviors under different scenarios, the combination of PBS and ABS can provide a better situational awareness capability by considering both the sensory inputs and human decisions. Furthermore, advances in sensory detection and tracking technologies allow for real-time planning and control of the surveillance system in the broader area. This paper aims to devise an optimal planning and control policy for surveillance systems, which will process different types of sensory data including videos, seismic data, as well as behavior models. To consider different scenarios in the surveillance area, we formulate this problem as a Markov Decision Process (MDP) by utilizing various sensory data for parameter selection. We then develop a Digital Twin (DT) of the surveillance using both PBS and ABS to calibrate and validate our proposed MDP framework. The resulting multi-paradigm simulation framework with DT can be an attractive approach to handle uncertainties in a system caused by the heterogeneity and velocity of the sensory data.