Mobile robot navigation is a constantly evolving field that is adopting new paradigms along the way, and recent methods, such as Transformer-based models, have helped facilitate advancements in perception and decision-making tasks in this decade alone. This paper explores modern scene understanding techniques, including Contrastive Language-Image Pretraining (CLIP) and its role in improving semantic scene comprehension for various indoor environments. Existing benchmarking methods for evaluating autonomous mobile robot navigation performance are limited in accommodating the dynamic nature of real-world scenarios. Therefore, a set of metrics is proposed for robust evaluation, highlighting the need for standardized frameworks that meet modern expectations. Furthermore, a multimodal robot navigation model is introduced; it consists of visual and laser data combined with semantic embeddings to augment navigation performance. The proposed model and metrics aim to contribute to better benchmarking standards for indoor robot navigation systems.