Super-Twisting Sliding Mode Control and Observation for a mobile ground robot

An Observer and a Controller based on the Sliding Mode theory have been developed for a mobile ground robot. After an introduction about the main control theory concepts, including classical control techniques such as Proportional Integral Derivative (PID), additional considerations about the utility and use of observers in practical cases are provided. The theory of sliding mode is presented and analyzed in terms of performance, stability and convergence. The mathematical stability of the control and observation algorithm is studied making use of Lyapunov theory. The Observer and Controller design is based on the kinematic model of the unicycle. Matlab and Simulink have been used as simulation environment in which a pre-existing vehicle model is present. An initial tuning in simulation has been performed for both the navigation and control blocks. For the experimental case, a practical trial-and-error phase has been addressed to assess performance issues, such as sensor noise, non-linearities, data sampling frequencies, making it possible to deal with such limitations. An overview of the nodes and topics already on the robot board has been addressed, including a description of the Extended Kalman Filter (EKF). The filter performance is compared with the new sliding mode super-twsting observer. The observer estimates the body velocity and orientation angle in an inertial reference frame using as inputs acceleration and angular velocity, from IMU sensors, and inertial position from encoders. The controller takes as input the inertial orientation and body velocity from the observer and a reference trajectory from a ROS node containing the Artificial Potential Field (APF) algorithm. The algorithm from simulation are translated to Python code to be deployed on the real Unmanned Ground Vehicle (UGV). The Observer is tested on the real robot, initially in an open loop configuration. Eventually, the Observer is implemented in feedback with the Controller. The influence of the nonlinear closed loop system on the robot dynamic is then discussed. The results of the thesis enhance the robustness of the Sliding Mode technique, showing optimal performances from both the Observer and Controller.