The　neural　network　feed-forward　control　of　free-floating　dual-arm　space　robot　system　in　joint　space　is　discussed　in　this　paper.　According　to　the　law　of　conservation　of　the　linear　and　angular　momentum,　the　dynamic　equations　of　space　robot　systems　are　established　through　Lagrange　equation　of　the　second　kind.　Based　on　the　above　results,　the　dynamic　equations　are　modeled　by　Gaussian　radial　basis　function　neural　network.　A　combined　scheme　of　neural　network　feed-forward　control　and　routine　feedback　control　for　space　robot　systems　with　unknown　parameters　is　proposed　to　track　desired　trajectories　in　joint　space.　The　proposed　control　scheme　needs　neither　linearly　parameterize　the　dynamic　equations　of　the　system,　nor　know　any　dynamic　parameters.　Besides,　it　uses　an　on-line　stable　weight　updating　mechanism,　so　it　does　not　require　the　time-consuming　training　process　and　saves　the　training　time　of　neural　network　To　show　the　performance　of　the　proposed　controller,　a　simulation　is　carried　out　on　a　planar　free-floating　dual-arm　space　robot　system.　The　simulation　results　show　that　the　proposed　control　scheme　is　feasible　and　effective.