The recent advancements in robotic systems have seen a significant shift towards integrating advanced vision-language models (VLMs) and large language models (LLMs) to enhance spatial reasoning, task planning, and real-time decision-making. A notable trend is the development of frameworks that leverage multimodal data, including semantic-topo-metric representations and geometric priors, to improve the robustness and adaptability of robotic navigation and manipulation tasks. These innovations are particularly evident in aerial navigation, where zero-shot learning and open-vocabulary capabilities are being explored to navigate complex environments. Additionally, there is a growing emphasis on self-supervised learning and continual learning approaches that allow robots to adapt to dynamic and unpredictable environments without extensive labeled data. The integration of these technologies not only enhances the precision and efficiency of robotic operations but also broadens their applicability across various domains, from construction safety to agricultural automation. Notably, the use of diffusion-based image generation techniques in visual servoing represents a novel approach to overcoming traditional limitations, enabling more versatile and adaptive robotic control. Overall, the field is moving towards more intelligent, context-aware, and adaptive robotic systems that can perform complex tasks in real-world scenarios with minimal human intervention.