Mobile　video　has　become　one　of　the　most　valuable　services　in　next-generation　heterogeneous　networks,　and　users＇　quality　of　experience　(QoE)　is　recognized　as　its　important　performance　metric.　In　this　paper,　we　propose　an　adaptive　bitrate　(ABR)　algorithm　to　achieve　improvement　of　the　timevarying　QoE　determinants　during　a　mobile　video　playback.　Since　watching　videos　will　bring　mobile　data　charges　to　the　users,　the　traffic　consumed　by　video　downloads　should　be　customized　to　a　user-specified　amount.　To　this　end,　we　first　analyze　the　real-world　traces,　and　find　the　key　factors　to　formulate　a　continuous　QoE　model.　Then,　we　introduce　a　traffic-Aware　rate　adaptation　strategy　(TARA).　Given　that　users　are　aware　of　their　real-Time　capabilities　when　watching　a　mobile　video,　TARA　enables　robust　ABR　process　which　can　satisfy　users＇　requirements　with　traffic　constraints.　Finally,　a　centralized　reinforcement　learning　(RL)　approach　is　proposed　for　the　joint　optimization　of　TARA,　which　aims　to　maximize　the　designed　QoE　metric,　and　deliver　the　expected　viewing　experience　to　users.　Results　of　experiments　driven　by　both　the　simulated,　and　real-world　network　traces　reveal　the　efficiency　of　the　proposed　TARA　strategy,　and　demonstrate　its　signicant　performance　improvement　as　compared　to　the　state-of-The-Art　ABR　algorithms.　©　2013　IEEE.