Reasons that cause computational difficulties in collaborative optimization are analyzed and the COMI（Collaborative Optimization Based on Multidisciplinary Inconsistency） algorithm is presented．Relax constraints and system level penalty function are constructed based on inconsistency information between disciplinaries, and a GA optimizer is used to solve a typical MDO question．Simulation results show that the feasibility and the optimal value of COMI is more balanced than the standard CO algorithm and the relax CO algorithm．