With　the　integration　of　the　advanced　Internet　of　Vehicles　and　autonomous　driving　technology,　connected　and　autonomous　vehicles　(CAVs)　possess　a　stronger　information　perception　ability,　real　-time　communication　and　cooperation　ability,　and　a　shorter　reaction　time.　CAVs　reveal　great　potential　in　increasing　traffic　efficiency　and　promoting　sustainable　development　of　urban　traffic　systems.　Accordingly,　the　introduction　of　CAVs　in　the　existing　traffic　system　not　only　changes　the　driving　behavior　of　vehicles　but　also　reshapes　the　spatial　distribution　of　traffic　flow.　To　measure　the　impact　of　CAVs　on　urban　traffic　systems　at　a　macro　level,　we　first　propose　the　concept　of　the　road　network　capacity　in　a　mixed　human-driven　vehicle　(HV)　and　CAV　environment　and　define　it　as　the　maximum　total　travel　demand　that　can　be　accommodated　by　the　road　network.　Two　nonlinear　programming　models　(NLP)　are　established　to　formulate　and　calculate　the　road　network　capacity　(RNC)　with　mixed　HV　and　CAV　flows　based　on　the　assumption　that　CAVs＇　route　choice　behavior　follows　the　UE　principle　and　the　system　optimal　(SO)　principle,　respectively.　Since　the　existence　of　the　equilibrium　conditions　makes　the　established　model　challenging　to　solve,　we　reformulated　the　proposed　model　as　a　mixed-integer　linear　programming　(MILP)　after　employing　a　piecewise　linear　approximation　approach　and　solved　it　with　the　commercial　solver.　Finally,　several　numerical　experiments　based　on　Nguyen-　Dupuis＇s　network　demonstrate　the　validity　of　the　proposed　models　and　solution　method.　The　change　in　the　RNC　with　the　variation　of　the　CAV　penetration　rate　and　the　reaction　time　of　CAVs　are　also　analyzed　by　conducting　a　set　of　sensitivity　experiments.