The project delves into the realms of reinforcement learning (RL) and machine learning (ML) systems, addressing the challenges these algorithms face in comparison to human problem-solving abilities. While humans instinctively employ intuitive mechanisms like filtering irrelevant data, breaking problems into manageable parts, and adapting problem-solving strategies, traditional RL and ML algorithms struggle to replicate such nuances. To bridge this gap, the project endeavors to enhance RL agents with human-like capabilities.
The core of the project involves the creation of a system that imparts these “human” insights to RL agents. Through a preliminary analysis of problems, the system constructs learning agents that implement Proximal Policy Optimization (PPO). This system then acts as an intermediary between said agents and their environment throughout their operation, enhancing the performance of PPO agents by translating their learning into superior outcomes compared to those without the mediation. testing in the RDDLgym environment, alongside self-defined challenges, substantiates the efficacy of the system. Results highlight improved learning process efficiency and solution quality, surpassing or at least matching an equivalent learning network. The project not only demonstrates the possibilities of combining RL and deterministic approaches to enhance learning systems but also opens avenues for further exploration.
Overall, the project introduces a compelling approach to augmenting RL agents with human-like problem-solving instincts using a deterministic problem processing, showcasing enhanced performance and offering valuable insights for future advancements in the field.