Intelligent manufacturing workshop is highly integrated with mechanical equipment, sensor equipment and other hardware, edge computing, control, data acquisition and operating system. However, at present, some traditional workshops mainly rely on manually driven forklifts to transport materials, which is inefficient. Automated Guided Vehicle (AGV) is a kind of material handling equipment, which has many advantages such as high degree of automation, flexible application, safety and reliability, unmanned operation and so on. In this paper, a step control algorithm is proposed to solve the NP problem in the path planning of multi-AGV systems and the problem that the system stability is reduced when the number of system path nodes increases. Introducing the concept Agent, by introducing the time threshold and the strategy of trust evaluation and buffer, the number of bids and the scope of bid selection in the traditional contract network protocol are limited. The effectiveness of the algorithm is verified by data experiments, and the key parameters are analyzed.