In　this　paper,　we　solve　the　formation　problem　of　nonholonomic　mobile　robots　(NMRs)　over　noisy　communication　networks.　The　communication　signals　with　noisy　information　are　exchanged　among　different　NMRs　by　a　directed　graph.　Only　the　individual　mobile　robot　directly　receives　the　actual　and　feasible　signals　itself　without　communication.　Besides,　the　sensor-to-control　signals　suffering　from　the　noise　are　modeled　by　a　nonlinear　function　including　unknown　parameters　and　uncertainties.　Combined　with　the　adaptive　control　technique　and　robust　control　technique,　the　unknown　parameters　and　uncertainties　of　noisy　signal　models　are　respectively　estimated　and　compensated,　respectively.　Then,　a　novel　distributed　adaptive　formation　controller　is　proposed　to　guarantee　convergence　to　the　optimal　positions　over　noisy　communication　networks.　It　is　proved　that　the　formation　errors　are　convergent　to　a　small　neighbourhood　of　origin　and　closed-loop　signals　of　each　NMRs　are　bounded.　A　simulation　example　is　given　to　demonstrate　the　theoretical　studies.　©　2022　IEEE.