As　the　requirements　on　powertrain　efficiency　of　electric　vehicles　(EVs)　are　increasing,　integrated　motor-transmission　(IMT)　powertrain　systems　for　EVs　are　becoming　a　promising　solution.　For　the　integration　of　IMT　powertrain　systems,　the　system　state　information　and　the　actuator　status　are　usually　required　for　the　closed-loop　controller　design　or　the　on-board　fault　diagnosis.　Embracing　the　demands,　an　observer　for　simultaneous　estimation　of　input　and　system　state　of　an　IMT　powertrain　system　is　studied　in　this　paper.　It　is　well-known　that　controller　area　network　(CAN)　has　been　dominant　in　the　vehicle　network,　which　is　used　to　communicate　among　controllers,　sensors,　and　actuators.　However,　the　CAN　bus　always　induces　time-varying　delays　when　there　are　a　number　of　communication　nodes　on　the　bus.　The　CAN-bus　induced　delay　would　result　in　vibrations　in　the　vehicle　powertrain　or　even　deterioration　of　the　entire　closed-loop　system.　To　deal　with　the　CAN-bus　induced　delay　in　the　estimation　work　for　IMT　powertrain　systems,　the　potential　random　delays　are　considered　in　a　three-state　nonlinear　model　which　represents　the　behavior　of　an　IMT　system.　To　estimate　the　input　and　state　simultaneously,　an　adaptive　unscented　Kalman　filter　(AUKF)　is　adopted.　As　we　know,　the　adopted　AUKF　has　the　benefits　of　dealing　with　system　nonlinearities　and　calculating　the　noise　covariance　matrix　automatically.　Simulations　and　comparisons　are　carried　out.　We　can　see　from　the　results　that　the　proposed　observer　estimates　the　input　and　system　state　well.　Moreover,　the　resulting　estimation　error　is　smaller　comparing　with　the　estimation　error　of　the　observer　based　on　extended　Kalman　filter　algorithm.　IEEE