Several rehabilitation approaches have shown that robot-assisted therapy (robot-AT) can improve the quality of upper limb movements in children with cerebral palsy (CP). However, there is still no method for assessing upper limb motor function impairment using a combination of surface electromyography (sEMG) and inertial measurement unit (IMU) sensors. The aim of this study was to develop a functional ability model to assess the effectiveness of robot-AT on improving upper limb function in children with CP. Fifteen healthy children and fifteen children with CP were included in this study. Children with CP performed eighteen robot-AT sessions and were evaluated twice, using EMG and three-axis IMU readings from accelerometer (IMU-ACC). Principal component analysis and the RELIEFF algorithm were used for dimensionality reduction of the feature space. The classification was performed by using support vector machines, linear discriminant analysis, and random forest. The proposed assessment method was evaluated by using leave-one-out cross validation. With this approach, it was possible to differentiate between healthy children and children with CP pre-robot-AT and post-robot-AT with an overall accuracy of 97.56%. This study suggests that there is potential for modeling the assessment of the upper limb motor function impairment in children with CP using sEMG and IMU-ACC sensors.