In this study, the thermal-responsive polymer networks based on poly(vinyl butyral) (PVB) are prepared, and their friction properties in response to external stimuli are investigated. Under dry sliding condition, the materials show low friction (COF similar to 0.14) at room temperature, but show ultra-high friction (COF similar to 1.09) at 100 A degrees C above the glass transition temperature of PVB. This marked variation is due to the effect of recovery stress caused by the shape memory effect of polymer networks. Additionally, the recovery stress would increase with the increase of cross-linked density and test temperature above T (g), leading to a higher COF. The polymer networks also show excellent mechanical strength with tensile modulus and elongation at break over 60 MPa and 100 %, respectively. To the best of our knowledge, this is the first paradigm about tunable friction properties realized by shape memory polymer. These interesting properties would enable the polymer networks with potential application in the design of intelligent device in future.