This　paper　addresses　robust　consensus　for　a　class　of　linear　multi-agent　systems　(MASs)　with　uncertainty　in　communication　network.　It　is　suppose　that　the　weighted　adjacency　matrix　of　the　MAS　is　known　to　lie　within　an　uncertainty　polytope.　Base　on　the　Lyapunov　stability　theory　combining　with　homogeneous　parameter-dependent　(HPD)　Lyapunov　functions,　necessary　and　sufficient　conditions　for　ensuring　robust　consensus　of　the　MAS　are　given.　Furthermore,　how　these　conditions　can　be　checked　through　convex　programming　is　provided　with　the　aid　of　sum　of　squares.　Numerical　examples　are　presented　to　illustrate　the　effectiveness　of　the　approach.