The　self-learning　control　of　a　free-floating　space　flexible　manipulator　based　on　fuzzy　neural　network　is　studied.　The　dynamic　equations　of　the　system　can　be　developed　by　using　the　Lagrangian　assumed　modal　methods.　Because　the　momentum　conservation　is　strictly　developed,　the　dynamics　of　the　space　flexible　manipulator　is　nonlinear　strongly.　When　the　fuzzy　control　and　neural　network　are　combined,　namely　the　fuzzy　inference　is　realized　by　neural　network,　the　fuzzy　control　can　self-learn.　Based　on　the　results,　the　self-learning　controller　of　the　space　flexible　manipulator　in　a　joint　space　is　designed.　The　controller　does　not　require　the　measuring　position,　the　velocity　as　well　as　the　acceleration　of　the　base.　And　the　numerical　simulation　is　carried　out,　which　confirms　the　controller　proposed　is　feasible　and　effective.