Uncertainty-Aware Continual Reinforcement Learning via PPO with Graph Representation Learning | Synapse