Automatic parcel sorting system using Automatic Guided Vehicle (AGV) is a research hotspot of intelligent logistics, in which path planning is one of the key issues. In an express sorting system, AGV has the characteristics of high density and large number of vehicles. This situation will cause congestion, making the performance of the whole system decrease. Aiming at this problem, this study proposes a CAA^* (Congestion-Avoidable A^*) algorithm that is able to avoid congestion. Based on A^* algorithm, the proposed algorithm introduces dynamic attribute nodes and establishes dynamic environment model to predict the possible crowding situation of each node, judge whether there are potential congestion nodes in the path planning process and bypass possible congestion nodes. Experimental results show that the proposed CAA^* path planning method can effectively reduce congestion in high-density and large-scale AGV scenarios, thereby improving the density of AGV and system sorting efficiency. Simulation results on practical application sites show that the proposed algorithm improves the AGV density by 28.57% and the sorting efficiency by 24.29% compared with the traditional A^* algorithm.