Artificial Intelligence in Collaborative and Industrial Robotics

Recent advances in artificial intelligence are reshaping collaborative and industrial robotics, enabling a transition from deterministic, pre-programmed automation toward adaptive, learning- enabled systems. This paper synthesises developments in imitation learning, diffusion-based visuomotor policies, and foundation models, and examines their integration within industrial robotic architectures. Particular attention is given to the convergence of language-based planning, multimodal perception, and digital twins for safe and flexible deployment. Electric vehicle battery recycling is considered as a representative high- variability and safety-critical case study, illustrating how contact-rich manipulation, sim-to-real transfer, and certified runtime supervision can be combined within a unified framework. It is argued that the same AI stack supporting flexible assembly in manufacturing can be extended to other related areas, such as disassembly-related circular-economy processes. Open challenges remain in safety certification, explainability, data scarcity, and multi-material interaction modelling. Future directions include cognitive digital twins, tactile foundation models, federated learning, and multi-robot coordination. The convergence of learning-based control and industrial digital infrastructures provides a pathway toward resilient and sustainable Industry 5.0 production systems.