PERTANIKA JOURNAL OF SCIENCE AND TECHNOLOGY

Reinforcement learning (RL) has achieved significant success in complex, sequential decision-making tasks. However, it remains constrained by its dependence on predefined reward functions, limiting adaptability in dynamic environments. Inverse Reinforcement Learning (IRL) addresses this limitation by inferring reward structures from expert demonstrations, enabling more flexible and context-aware agents. The study explores the potential of IRL’s in enhancing the efficiency and adaptability of modern autonomous systems. The pivotal role of IRL in modelling human-like reasoning and imagination is examined across domains, including robotics, autonomous driving, personalised medicine, and cybersecurity, alongside discussion on current solutions, challenges, and emerging research directions. The findings underscore future improvements for human cognitive capabilities and machine autonomy.