Abstract: In a matrix-arranged job shop, the finished workpieces are transported via automated guided vehicles (AGVs) from processing stations to downstream material supermarkets. This process needs to satisfy extremely tight time constraints of downstream delivery to stabilize production flow, reduce work-in-process inventory, and minimize transportation costs. To this end, a multi-AGV scheduling model is formulated and an improved multi-objective migrating bird optimization algorithm is proposed with the objectives of maximizing on-time delivery ratios and minimizing transportation costs. Specifically, objective-oriented delivery time determination and AGV allocation rules based on the earliest delivery deadline are proposed to achieve feasible decoding with no delay and minimal inventory. A Pareto front advancement mechanism is facilitated through a greedy operation combined with an ideal-point-based local search operator. A two-point crossover operator and an acceptance criterion of inferior solutions considering high-value information are proposed to ensure the diversity and convergence of the Pareto solution set. Experimental results indicate that the proposed algorithm can solve the problem effectively, achieving superior performance in terms of IGD and HVR indicators for the obtained Pareto frontiers.