This　paper　investigates　the　robust　mean　square　consensus　problem　for　linear　discrete-time　multi-agent　systems　over　arbitrarily　switching　topologies　with　unknown　transition　probabilities.　The　switching　is　governed　by　a　time-homogeneous　Markov　process　from　a　set　of　possible　interaction　topologies　among　agents.　A　sufficient　condition　is　proposed　for　achieving　mean-square　consensus　under　a　common　control　protocol.　Moreover,　the　solution　of　the　controller　is　formulated　as　a　group　of　LMI　constraints.　Finally,　the　theoretical　results　are　validated　via　a　numerical　simulation　example.　©　2019　Technical　Committee　on　Control　Theory,　Chinese　Association　of　Automation.