In the thesis we set up metrics that describes the motion quality of the mobile robots. We split those metrics into two groups. One that describes the position accuracy of the robot, so how much the robot can approach the target position, and one that describes how much the robot can follow the commanded trajectory. In general the accuracy of localisation of the robot has a strong impact on the robots motion, but in this thesis we only focus on the quality of the motion, not the precision of the localisation. After setting those metrics up we attempt to improve the motion quality of the KUKA holonomic robots. For that we first create a dynamic model about the holonomic robot with the help of the Lagrangian equation and we providing a process that we can extract the parameters of this dynamic model with. After that we introduce the feed-forward torque control that we use to improve the motion quality of the robot and we implement and integrate it to the KUKA OS system. For the end we make a measurement to investigate the effect of the feed-forward control on the motion quality.