Design and comparison of controller performance on four mecanum wheeled mobile robot

In this thesis the main aim is to design and compare controller performance on omnidirectional mobile robot based on mecanum wheels and it can be use for further researches for elemination system errors. On the purpose of to increase orientation capability the system is equipped with four mecanum wheels which have twelve rollers around it. In the scope of this thesis to control dc motor speeds according to user demans and to minimize heading errors in the band of accaptable area. Before start to develop controller for the large scale system experimental test rig has been equipped with system components which are mecanum wheels, dc motors, encoders and controller. Also before start design controller all system elements should have mathematical models to take best controller results. Such as Dc motor model, system forward and backward kinematics, which help to minimize heading errors in real nonlinear environment. For the developing linear quadratic regulator and model predictive controller transfer function of the system is needed. Especially for LQR controller owing to feedback constants. The dc motor model is implemented to the simulation enviroment to see system output according to user inputs. In order to validate designed model, different system scenarios are tested independetly. After validation of system outputs cross validation test is conducted to maximize dependability. Designed model simulated in Matlab/Simulink environment with real dc motor coefficients. To improve controller performance for LQR controller different Q matrixes and R constants are tested and simulated. In the end admissible feedback constants are chosen. For MPC system inputs and outputs are simulated with modelled system. The controller performance is adjusted according to time interval, control horizon, weights of inputs and prediction horizon constants. Based on simulated model the aim is to take smooth response from system without overshoot, minimal oscillation and minimize settling time. Simulated results will be compared to real system results in future.