ROS Gazebo and MATLAB/Simulink Co-simulation for Cart-Pole System: A Framework for Design Optimization

Co-simulation is a process of combining several simulation environments simultaneously. This approach is critical in requirements to model the interactions of multiple physical phenomena or disciplines. This paper presents in depth an integrated co-simulation framework of Robot Operating System (ROS) and MATLAB/Simulink for the Cart-Pole system. Cart-Pole is known as one of the classic experimental systems of control theory. This research firstly discusses the mathematical modelling of the Cart-Pole system in detail and explains how this model can be simulated linearly and nonlinearly in MATLAB. It also provides information on how to implement the LQR control structure for the Cart-Pole example in both linear and nonlinear simulations in MATLAB and in the Gazebo simulation. The research introduces the integration process between ROS-Gazebo and MATLAB/Simulink (Cosimulation framework) as well as the tools and libraries used. In the context of design optimisation, it is shown how to integrate this co-simulation framework with evolutional algorithms. In conclusion, this paper shows that MATLAB-linear, MATLAB-nonlinear and ROS-Gazebo simulations produce different results and comments on the reasons for these differences. For researchers seeking design optimisation through Reinforcement Learning (RL) approaches, the framework allows for simple simulation of dynamically modelling highly complex problems.