Abstract

Robot simulators are valuable tools for researchers to develop control code in a fast and efficient manner without spending time setting up physical experiments. Most simulators, however, do not model the real world accurately. As a consequence, when a program is run on a real robot it may behave differently from when run in simulation. In this paper we present a method of developing a robot simulator that models the operation of a real robot in a real environment accurately, using real robot data and system identification to construct the simulator’s model.