Shifting from model-free to model-based RL approaches.
Implement a deep-world algorithm (e.g., a simplified Dreamer variant) and compare with model-free baselines.
GeeksforGeeks blog on model-based reinforcement learning
Shallow world models for continual RL