Robotics at a Crossroads: AI-Based vs Classical Methods in Control, HRI, and Autonomy

The field of robotics is undergoing a profound transformation driven by rapid advances in artificial intelligence, particularly large language models and vision-language models. These AI-based approaches are increasingly being integrated into robotic systems for control, decision-making, perception, and human-robot interaction, often demonstrating remarkable capabilities in generalisation and adaptability. At the same time, classical methods—rooted in well-established control theory, model-based planning, and rule-based systems—continue to offer reliability, interpretability, and formal guarantees that remain essential in safety-critical applications. This evolving landscape raises fundamental questions about when and how AI-driven approaches should complement or replace traditional techniques, and whether hybrid models combining classical and AI-based methods can leverage the strengths of both paradigms. Furthermore, while established metrics and evaluation frameworks exist for classical methods, the evaluation of AI-based systems—particularly those built on large language and vision-language models—remains an open challenge requiring new methodologies and benchmarks.This Research Topic aims to address the critical need for systematic investigation into the comparative performance, integration, and evaluation of AI-based and classical methods in robotics. As large language models, vision-language models, and reinforcement learning techniques become increasingly prevalent in robotic control, human-robot interaction, and autonomous systems, the research community faces pressing questions: Under what conditions do AI-based methods outperform classical approaches? How can we develop robust evaluation frameworks for AI-driven robotic systems? And how can hybrid architectures effectively combine the adaptability of AI with the reliability of classical methods? Recent advances in foundation models have enabled robots to interpret natural language commands, reason about complex environments, and learn from limited demonstrations. However, challenges remain regarding safety, explainability, real-time performance, and generalisation to novel scenarios. This Research Topic seeks to bring together studies that develop, evaluate, and compare AI-based and classical methods, propose novel hybrid approaches, and establish new benchmarks and evaluation methodologies. By fostering this dialogue, we aim to guide the robotics community toward principled decisions about when and how to deploy these different paradigms.This Research Topic welcomes original contributions addressing the development, application, and comparative analysis of AI-based and classical methods in robotics. Specific themes include: AI and LLM-based robotic control strategies; vision-language models for robot perception and action; classical control theory and model-based approaches; hybrid architectures integrating AI and classical methods; reinforcement learning for autonomous systems; human-robot interaction using natural language interfaces; cognitive robotics and decision-making systems; evaluation frameworks and benchmarks for AI-driven robots; and safety, explainability, and reliability in AI-based robotic systems. We invite Original Research articles presenting novel methods or experimental findings, Review articles synthesising current advances, Methods articles proposing new evaluation frameworks or benchmarks, and Perspective articles discussing future directions and open challenges. Comparative studies that rigorously evaluate AI-based versus classical approaches across specific applications are particularly encouraged. Topic editors: Abolfazl Zaraki, Hamed Rahimi Nohooji, Yeganeh Madadi Source: https://www.frontiersin.org/research-topics/78681/robotics-at-a-crossroads-ai-based-vs-classical-methods-in-control-hri-and-autonomy