This paper examines the integration of behavior primitives into reinforcement learning for robotic manipulation. Primitives break tasks into structured actions, enhancing learning efficiency, adaptability, and generalization. A hierarchical approach improves decision-making, while affordance-based exploration optimizes learning. Key challenges include designing effective primitives and ensuring real-world robustness.

Альшауа Р. (науч. рук. Борисов И.И.) Reinforcement learning with behavior primitives for manipulation tasks // Сборник тезисов докладов конгресса молодых ученых. Электронное издание. – СПб: Университет ИТМО, [2025]. URL: https://kmu.itmo.ru/digests/article/15372