Recent advances in large language models (LLMs) are transforming artificial intelligence from passive language processing systems into autonomous agents capable of reasoning, planning, tool use, communication, and interaction with physical environments. At the same time, progress in multi-agent systems, multi-agent reinforcement learning (MARL), embodied and physical AI, robotics, and edge computing is enabling intelligent agents to collaborate and operate autonomously in increasingly complex real-world environments. Technologies such as retrieval-augmented generation, network architecture search, hyperparameter optimization, and adversarial machine learning are included insofar as they support the design, implementation, robustness, or evaluation of these agentic and autonomous systems.

This Topical Collection aims to bridge the gap between LLM-based intelligence and physically deployed autonomous systems by bringing together research on agentic AI, autonomous and multi-agent systems, robotics, embodied and physical AI, and edge computing. It welcomes Research Articles, Method Articles, and Review Articles that present original research, emerging findings, reproducible methods, system implementations, evaluation frameworks, or critical syntheses of the field.

Contributions may address models, prompts, retrieval and knowledge integration, communication protocols, learning and optimization methods, simulation environments, hardware platforms, sensors, safety and robustness, deployment procedures, and evaluation, provided that these elements are clearly connected to the implementation or study of agentic, autonomous, multi-agent, embodied, or physical AI. Through original studies, emerging results, methodological advances, and reviews, the collection will help researchers understand, reproduce, compare, and extend approaches that move from LLMs to autonomous agents, from individual agents to collaborative intelligence, and from digital intelligence to real-world physical AI.