Research on Dynamic Path Planning of Multi-AGVs Based on Reinforcement Learning

Automatic guided vehicles have become an important part of transporting goods in dynamic environments, and how to design efficient path planning method for multiple AGVs is a current research hotspot. Due the complex road conditions there may be obstacles situations which only target point known but complete map lacking, leads poor long time automatic (AGVs). In this paper, two-level (referred as GA-KL, genetic KL method) multi-AGVs proposed by integrating scheduling policy into global combining algorithm local algorithm. First, planning, we propose improved Q-learning optimization (K-L, Kohonen algorithm) based on network, can avoid autonomous finding using learning function Then, adopt idea K-L with algorithm; addition, integrate continuously adjust according changes environment. Finally, through simulation field experiments, verified that accomplish finding; compared traditional algorithm, achieved improves results length convergence various maps; was reduced about 6.3%, average, 4.6%, average. The experiments also show GA-KL has satisfactory search capability effectively obstacles. final demonstrated total completion average 12.6% 8.4% narrow working environments or highly congested situations, considerably efficiency multi-AGVs.