Mobile　cloud　computing,　which　comes　up　in　recent　years,　is　a　new　computing　paradigm.　In　mobile　cloud,　mobile　users　can　access　and　schedule　the　resources　or　services　in　remote　clouds　via　wireless　networks,　which　we　call　mobile　cloud　task　scheduling.　They　even　can　build　mobile　micro-cloud　(MuCloud)　with　mobile　device　to　provide　lightweight　service.　However,　unreliable　wireless　connection　and　dynamic　join　and　quit　of　MuCloud　make　task　scheduling　in　mobile　cloud　face　more　challenges　than　in　wired　cloud.　Moreover,　from　both　the　users　and　service　providers＇　perspective,　task　scheduling　is　a　multi-objective　optimization　problem.　Small　makespan　and　load　balancing　are　pursued　by　mobile　users　and　cloud　service　providers　respectively.　In　this　paper,　we　advance　a　demand-driven　task　scheduling　model　and　introduce　an　estimate　method　to　predict　warranty　complete　time　of　tasks　in　wireless　network.　An　improved　genetic　algorithm　using　2D　chromosome　(2DCGA)　is　presented　to　tackle　multi-objective　task　scheduling.　Simulation　experiments　show:　1)　compared　with　Markov　model,　our　estimate　method　has　higher　accuracy　of　prediction　and　more　reasonable　prediction　results　of　probability　of　task　scheduling　failure;　2)　2DCGA　has　good　performance　for　task　scheduling.　When　compared　with　IGA,　it　has　smaller　makespan　and　lower　deviation　of　load;　3)　objective　priority　can　be　adjusted　exactly　by　weights　of　fitness　functions.　It　makes　2DCGA　suitable　for　multi-objective　optimization.　©　2014　IEEE.