Fuzzy Clustering-Induced Switching Reinforcement Learning with Deterministic Annealing

Key Points

• The proposed model achieves more efficient Q-learning by leveraging fuzzy clustering techniques with multiple agents.
• Fuzzy c-Means is utilized to compute memberships, improving Q-value updates for parallel learning in distinct environments.
• By incorporating deterministic annealing, the model enhances robustness and maximizes the gains of agents during learning.
• This approach may enable better adaptability of agents to varying environmental complexities, optimizing reinforcement learning outcomes.

Abstract