The integration of IoT devices will play a key role in linking cyber and physical spaces. Combined with Digital Twin, these devices are mandatory for the development of the Metaverse. They will provide it with a nervous system by collecting, transforming and distributing the data in real time, generating a Dynamic Digital Twin. However, the cost of deploying IoT networks is a major barrier to the growth of this new technology. Optimizing this deployment is tedious particularly in complex environments, as the devices require connectivity, which is a constraint to be taken into account in addition to the objective of maximizing coverage. In this paper, we propose a new two-step method named PREP-NSGA to optimize IoT indoor deployments considering connectivity constraints, attenuation of the coverage, and heterogeneous obstacles whose information is extracted from the Digital Twin database. The first step of PREP-NSGA consists of a preprocessing based on the skeleton extraction of the structure, while the second step is a pareto-based multi-objective genetic algorithm. The performance evaluation is conducted using numerical and simulation tools on both real and computer-generated scenarios. The obtained results show that the proposed solution PREP-NSGA outperforms existing solutions and random deployments in terms of coverage within connectivity constraint.