Oncolytic virus therapy aims to treat cancer using viruses. This promising therapy is weakened by clinical challenges related to dosage, toxicity, and uncertain tumor dynamics. The interactions between oncolytic viruses and tumor cells are described by mathematical models to advance understanding of treatment outcomes and design better therapies. Motivated by the urgent need to improve clinical outcomes, this study proposes a nonlinear estimation and control scheme to determine doses of viral injections by applying impulsive control theory. The scheme is designed by using the extended Kalman filter and nonlinear model predictive control (MPC). The MPC relies on a model of oncolytic adenovirus therapy, where the model parameters have been identified by fitting the model to experimental data from five nude mice. The proposed scheme determines optimal viral doses, which tend to enforce better and faster tumor regression than former protocols. Moreover, this scheme delivers personalized therapy which is robust to some parameter and modeling uncertainty. Together, the findings of this paper stand as an in-silico proof-of-concept to develop control engineering approaches, which help to resolve clinical challenges of oncolytic virus therapies.