Fig. 4 | Scientific Reports

Fig. 4

From: A hybrid APF-DQN framework with transformer-based current prediction for USV path planning in dynamic ocean environments

Fig. 4

Overall architecture of the proposed APF-DQN framework. The system comprises three main modules: (1) the Transformer-based ocean current predictor that forecasts local flow fields (Section “Multi-Scale Transformer Model for Ocean Current Prediction”); (2) the Enhanced APF module that computes physics-based navigation guidance (Section “Enhanced Artificial Potential Field”); and (3) the DQN module that learns optimal policies through state representation, reward shaping, exploration optimization, and policy fusion (Sections “State Space Representation”, “Reward Function”, “Deep Q-Network with Median-Based Exploration”, and “APF-Guided Deep Q-Network Integration”). Arrows indicate information flow between components.

Back to article page