In　this　paper,　the　kinematics　and　dynamics　of　free-floating　space　robot　system　with　dual-arms　are　analyzed.　The　closed-loop　dynamic　error　equation　considering　uncertainties　for　the　space　robot　is　derived　by　using　the　PD　type　computed　torque　method.　A　new　learning　control　method　using　real　coded　genetic　algorithm　is　presented　to　control　the　system　and　its　controller　structure　is　designed.　A　compensating　learning　control　item　is　added　to　the　error　equation　to　approach　the　uncertainties　by　evolutionary　learning.　Then,　the　computed　torque　method　and　the　GA　learning　control　are　combined　to　realize　the　perfect　trajectory　tracking.　Simulation　results　show　the　efficiency　of　the　control　method.