Nowadays, indoor localization is among the most important challenges in IoT networks. On the other hand, Deep Learning techniques are emerging as a leading method. Additionally, meta-heuristic algorithms attract several research domains due to its efficiency in resolving optimization problems. In this work, a Deep Learning model optimized by meta-heuristic algorithms and based on Time of Flight (ToF) measurements captured by Ultra-Wide Band technology, as an indoor IoT localization solution, is proposed. The findings showed that optimization with the Grey Wolf Optimizer can accelerate convergence towards optimal parameters during the learning phase. Furthermore, it was observed that ToF measurements enhance the positioning estimation capabilities, in comparison with RSSI and Range measurements. Compared with an existing multi-lateration method, the suggested solution provided more accurate positions of the IoT mobile object as it yielded better results in terms of localization accuracy (98.92%), Mean Absolute Error (0.057m) and Mean Squared Error (0.0095m).