Abstract: The multi-depot collaborative distribution vehicle routing problem with multi-commodity (MDCDVRPMC) was proposed to address the limitation of existing multi-depot vehicle routing research that primarily focused on homogeneous commodity distribution, by considering both depot inventory differences and customer multi-commodity demands. A mixed-integer programming model was established with the objective of minimizing transportation costs through order splitting to handle heterogeneous demands, and an enhanced adaptive large neighborhood search (EALNS) algorithm was designed for solution. The algorithm integrated K-means clustering, the savings algorithm, and greedy recombination strategies to generate initial solutions, while adaptive large neighborhood search was employed to prevent premature convergence, combined with 2-OPT neighborhood operations and the Metropolis criterion of simulated annealing for deep optimization. Finally, the model correctness and algorithm performance were verified through standard case tests using the Gurobi solver, adaptive large neighborhood search (ALNS), genetic algorithm (GA), and ant colony optimization (ACO). The results demonstrate that the EALNS achieves significant improvement in solving efficiency while ensuring solution quality (the solving time is only 2% of Gurobi's); compared with benchmark algorithms, the solution quality is improved by 13%-35% and solution stability is enhanced by 20%-40%, showing superior convergence performance and robustness. The research findings are recognized as providing an efficient solution for collaborative distribution in complex multi-depot logistics environments, effectively expanding the application scope of vehicle routing optimization theory in practical logistics scenarios.