Fig. 1. Flowchart of algorithm parameter update

Fig. 2. Framework of VLS navigation algorithm

Fig. 3. Collision risk capsule

Fig. 4. Model training scenario. (a) Simulation experimental environment; (b) Agilex Tracer

Fig. 5. Comparison of algorithms during the training phase

Fig. 6. Reward value distribution in the last 100 episodes

Fig. 7. GAZEBO experimental environment. (a) Simple scenario; (b) complex scenario

Fig. 8. Attention visualization. (a) Scenario 1; (b) scenario 2

Fig. 9. AGV experimental platform and experimental environment

Fig. 10. Qualitative algorithm and comparative experiment