Large-scale　multi-robot　task　allocation　(MRTA)　problem　is　an　important　part　of　intelligent　logistics　scheduling.　And　the　load　capacity　of　robot　and　picking　station　are　important　factors　affecting　the　MRTA　problem.　In　this　paper,　the　MRTA　problem　is　built　as　a　many-objective　optimization　model　with　four　objectives,　which　takes　the　load　capacity　of　single　robot,　single　picking　station,　all　robots　and　all　picking　stations　into　account.　To　solve　the　model,　a　hybrid　many-objective　competitive　swarm　optimization　(HMaCSO)　algorithm　is　designed.　The　novel　selection　method　employing　two　different　measurement　mechanisms　will　form　the　mating　selection　operation.　Then　the　population　will　be　updated　by　employing　the　competitive　swarm　optimization　strategy.　Meanwhile,　the　environment　selection　will　play　a　role　in　choosing　the　excellent　solution.　To　prove　the　superiority　of　our　approach,　there　are　two　series　of　experiments　are　carried　out.　On　the　one　hand,　our　approach　is　compared　with　other　five　famous　many-objective　algorithms　on　benchmark　problem.　On　the　other　hand,　the　involved　algorithms　are　applied　in　solving　large-scale　MRTA　problem.　Simulation　results　prove　that　the　performance　of　our　approach　is　superior　than　other　algorithms.